Transcription
Welcome to Better System Trader, the show for systematic and algorithmic traders. Uh, I've got a fantastic guest for you today, someone who has had a massive contribution to the technical analysis community. He's even won some awards for his research. Welcome Alex Spirago.
Welcome, welcome. Thank you so much, Andrew. Thank you so much for, uh, having me on your podcast. Uh, and I must say, I've been a fan of your show for a long time, especially, especially during COVID. You did manage to give me quite, you know, a lot of company because I had downloaded a lot of your, your interviews and I was listening to them. And, uh, thank you so much. It's, it's really nice to be here.
Oh, great. Well, it's an honor to have you on the show. Finally, I, I can't believe I didn't discover you much sooner. So I must have been hiding under a rock because you've, uh, you've published a lot of research and I see on LinkedIn, your epic, you, you publish so many things and you've got a l, a big following. So maybe I should hang out on LinkedIn more and discover fantastic people like yourself. But, uh, thanks for joining us today. Um, you're in the UK, so it's the evening, I understand. It's getting very late for you. So hopefully, uh, you know, it's early for me. Hopefully, we can figure this out together.
Right. I'm sure we will. I'm sure we will. Okay, so how about, just to get started, do you want to share a little bit of background on yourself? How you got into trading?
Absolutely. Well, it's, um, it started a long time ago. Well, my interest in markets, not trading itself, at a very young age. Well, um, um, my father was working for Bank of America. So one day he took a trip to England for a seminar. He went to Bank of America, the dealing room, the trading room. He saw it, and then he came back and with, uh, within his suitcases also came a lot of advice. Son, I think this is the best thing for you to go into a trading room and dealing room. As a young, 15-year-old child who wanted to impress their father, I said yes, without even knowing what trading is at the young age of 15. So, um, yeah, I did study, um, then afterwards in City University, City of London in England, did bank and finance. And, um, then I started, you know, because I had an interest in finance, but I didn't know much about capital markets, then I started crystallizing my, um, my view on capital markets. So, and at the last, so that was the first big, you know, uh, influence on my life, my father. The second one was at the last year of my studies, um, I took an elective course which really changed my life around. Um, so the course was called at that time, forecasting. It was taught by someone who later became a very good friend, Professor Roy Bachelor, who actually retired this year. And, um, basically, Roy was an ex-life trader. But, um, so he did use technical analysis. So, but he was, um, he, he didn't want to, you know, let's put another way. At that time, when I said at that time, talking about 1995, technical analysis, yeah, technical analysis wasn't as accepted in the academic community. So he basically did some econometrics at the back, some econometrics at the end, put the technical analysis in between. He made it into an econometrics or forecasting course. So I took that course. That was my first, um, uh, contact with, uh, technical analysis. I think I was 21, 21, 22 around that age. So that's when I learned technical analysis and I fell immediately in love with it. Then I did my military service, finished university, then home, worked in, uh, various investment management companies, proprietary trading houses, hedge funds, and then I kind of like retired to for to trade just purely my own account. Um, in the meantime, of course, lots of things happened, but that's the short story of my, short story of my life. I also run the UK chapter of the CMT Association. Yes. Met a lot of interesting people. Um, so these are the, uh, the very short version without trying to bore people with my personal details.
Yeah. Well, as I mentioned at the start there, you've won awards for your, um, your research. And this is actually the way I discovered you, just a few weeks ago. So I don't know where I've, I've been hiding, how I didn't discover you earlier. But, uh, last, um, last show I had Rob Hanner on the, um, as a guest, and he, he just won the, the name, I guess, 2023, um, award. I think it's, is it called the Founders Award?
Yes, absolutely.
And you actually won the year before that, 2022, and also the CMT Association Charles D Award. Sorry, I'm just checking my notes to make sure I get these names right. And the paper was called MACD V, Volatility Normalized MACD. Um, which I think is very interesting because I guess MACD and also, and also just like standard moving averages are very, very popular indicators, perhaps maybe the most popular for a lot of different styles of trading. Um, so how did you get into, uh, the MACD and, and working out, um, you know, how to improve it?
Right. Gosh, that's a good question. Um, well, my quest started around, I would say, probably 2014. Well, I was, first of all, I was using, uh, um, momentum indicators at that time. But, uh, there, if like, there are literally hundreds of momentum indicators that there are. I think it's, it's probably safe to say they are probably, uh, more indicators than traders. There are literally thousands of indicators. So in order for me to pick one, because I was always experimenting and as a trader was really into to technical analysis, what I was trying to do is open up the hood, if you will, and look at the math and trying to understand the quirks, the pros, and the cons of each one. So what I did basically is divide them into broad categories. Like if you, if you do a momentum indicator taxonomy table, you, in, you basically divide them in two parts. The first one is range-bound, uh, indicators. Like they go from zero to 100. They also have a normalized scaling. Uh, so range-bound and normalized. And the other one is, uh, unbound indicators. So I looked. So that was the first thing I did. Then I started using, uh, indicators from each category. So for the range-bound indicators, could be like RSI, Percent B, Percent R, although strictly speaking, the stochastics. Um, and from the other categories, like MACD, which is an unbound indicator, or the rate of change, and, and so on and so forth. So each one, each category has its plus and minuses. And as I started using them, I was happy in some situations to use one type of indicator and in some categories, in some other cases, I was happy to use another type of indicator. So around 2014, my quest decided to try to use, uh, a hybrid one would be the best of two worlds. So that's how it's all started.
Yeah. Yeah. Well, before we get into that, actually, and go into that into more detail, maybe we could just spend one minute or so talking about the basics, so that people, everyone's on the same page. So can you explain what is the MACD and how is it commonly used?
Perfect. All right. So the MACD is, uh, as you know, developed by, uh, Gerald Appel. Y, uh, it was first featured in one of his books. It later became, became one of the most popular technical analysis indicators. I think it's by a, a study done by Bloomberg was the most popular after RSI, right? So the, the simple, um, version of it is, it's constructed by using two moving averages, taking the difference between these two moving averages and plotting the difference. That's the MACD. So the, in, in, so the moving average in question was the 12 and 26 period lookbacks. And the principle of it is very simple. Since you have two different moving averages, one shorter, one longer, uh, when price, for example, um, rallies to the to the upside, the short-term moving average will react quicker than the longer-term moving average. So the spread between these two moving averages will increase as there is a sufficient momentum either to the upside and to the downside. And so when you see the spread increasing, you know you've got upside momentum. When it's decreasing, you know you've go, uh, after point, you've got downside momentum. And since it's an absolute price indicator, uh, it has its plus and minuses. I mean, the plus is that it can expand as much, uh, as price does. The minus is that it's an absolute price indicator. Mean, the absolute value itself is almost worthless.
Okay, so let's dig into that a little bit more, because I know in your research paper, you went into quite a lot of detail about some of the challenges of MACD. And, uh, to be honest, I didn't know about a lot of those challenges. Uh, so maybe there are other traders as well who aren't really sure about it. Uh, maybe can you explain what are some of the challenges with?
Sure, sure, sure. Um, I think, uh, let me, uh, for people to visualize, I've got an, uh, I think I sent a slide called Indicator Evaluation Scorecard. I don't know if it's, probably, I think it's slide number three, I think it's called. Yeah, that's the one. Yeah, perfect. So where I did, okay, so I've got range-bound indicators and I've got unbound indicators. Now, on the left part of the chart, I've got some, some criteria for us to objectively evaluate and compare and contrast the two broad categories of indicators. So when I talk about range-bound indicators, the same thing applies for RSI, Percent B, Percent R, stochastic, so so forth. And for unbound, the same thing for MACD, all the rest of it. So what it is, is try to see the effects of having a normalized scaling. So very quickly, uh, our indicator value is comparable across time. So, for example, for the RSI, the range-bound indicator category, yes, it is because it's, since it's in a zero to 100 range, um, so if you have like say 60% value, it is comparable across time. So if you go back in the 70s, you go back to the 80s, the 60% can be comparable because it's on a percent basis. So that's an advantage they have. Uh, the indicator values are comparable also across securities because it's, it's a comparable, uh, it's at a percent basis, it's normalized. So a 60% value on the S&P 500 is the equivalent of the same for the Euro. And since it's normalized across time and across securities, you can create, uh, what I call a, a life cycle model. Basically, you can objectively define overbought, oversold, um, you know, have it high, low figures when they can be comparable. For example, you can say, okay, if the market is oversold, so-called oversold, I say below 30, then I'll be taking buy signals. So you can use it across time and securities. So that's the advantage that range-bound indicators have. Now, the disadvantages they have is that the indicator values do not attract to trend. And that means that, for example, if there is a prolonged, this is all part of a presentation, I'll have to be using my hands to explain what I have the chart. So adapting to trend basically means that if a market strongly rallies, the indicator does not adapt to trend in the sense that its values being, uh, pegged from a zero to 100 scaling, it does, it can only go say within this range. So it cannot adapt to trends. For example, if the market continues rallying, the indicator will stop at these levels and then continues rallying, it would stop at these levels. Basically, the, um, value that you have for 70 or 70% speaking about the the RSI is not indicative of the absolute value of the market, the momentum it has. So the market can go from a thousand to 1500, the RSI will go very high, and then it can triple, and then the RSI will still be, you know, will go very high, but it will still be like 70 or 80%. So you won't get a feel for how really strong the market is. Uh, also, um, gets, uh, doesn't adapt to extreme readings. So basically, sometimes the market, you can see if it strongly rallies very strongly, the indicator gets at very high values and stays at these values. So this is a quite newbie mistake when they start using oscillators. They say, oh, the market is overbought, and the market stays so-called overbought for a prolonged period of time. Well, actually, it's not overbought, just very, very strong. But you cannot understand the difference because, you know, the, it's, it's a limited, uh, scale of values it can have. And because of that, it's not freely, um, able to to range as much as the market, adapt with the market. The diverging signals you get are not that high quality, uh, because it cannot truly adapt, it just gets pegged. So the market rallies, and the indicator sort of like flat lines, so you get this false sense of divergence. So these are the range-bound indicators, very quickly review plus and minuses. For the unbound indicators, you have the reverse situation. So where, uh, range-bound indicators shine, this is where they fail. For example, they're not comparable across time. For example, the MACD, because, uh, it's, the, the MACD is basically the, uh, secondary derivative of price. So you've got price, then you've got one moving average, and two moving averages, and then you take the difference between this. So it's the second moving average, the second derivative of price. The thing is, um, it's not comparable across time because it gives you an absolute value, an absolute price, and that is depending on the absolute price of the market in question. It's not comparable across time if there are huge fluctuations across time. So, for example, the S&P 500, because it rises, uh, has an upward bias, an upward drift. So for the last, you know, 50 years, because it keeps increasing, the absolute value of the MACD doesn't reflect absolute momentum, just it just reflects the absolute value of the S&P 500. So if, um, the S&P 500 for some reason went up to 50,000 points, the MACD value would just reflect the increased value of the MACD, will increase, will reflect the, um, the increased value of the S&P 500, the underlying security, not more momentum. So for that reason, it's also comparable across securities, which means you've got the MACD of S&P 500, which is measured in 10,000s of points, and then you've got, uh, a market low-priced market, so say currency, Euro, which is like four decimals. Um, so the differences between these two moving averages will, the absolute value will affect the absolute value of the underlying market, not momentum. So it's not comparable across time. So consequently, if it's not comparable across time and across securities, you cannot create a, a life cycle model when you say overbought, oversold, and so on and so forth. Right. So these are the minuses. So, and finally, reaching the advantages they have, the MACD itself, or rate of change, or any other kind of indicator, it can adapt to trend because since it's not normalized, as the market, market increases, it can adapt with it. So it, it can truly adapt to trends. It doesn't get pegged at high values. Also, it doesn't get extreme readings because as the market goes, so it does go. So it doesn't flat line. And thus, consequently, it has better divergent signals. So to conclude the scorecard in the indicator evaluation, you've got the normalized scaling with the range-bound indicators shine, and the trend influence where unbound indicators are better. And, uh, you've got the pros and cons, uh, completely, um, opposite to each other. So as I mentioned, 2014, when I, you know, created this like sort of evaluation, big eye view of how indicators are classified and where they work and they don't, my idea was, how can I create an indicator which is the best of two worlds? And the best of two worlds, I mean, take the normalized scaling of range-bound indicators, which is objective, compared across time, securities, and so on and so forth, but also have an unbound indicator which doesn't get have the limitations of range-bound. So have an, an unbound normalized indicator. So that was my, my, uh, my point of departure.
Right. Okay. So many traders, I think that was, um, that was a really good slide. Thanks for showing that, the looking at the pros and cons of indicators. Uh, now many traders would go, okay, well, I'll just have an unbound and I'll have a, a bound indicator and I'll switch between them, uh, when I have to. What do you think about that approach?
Well, it's certainly doable, but it's not very helpful because you'll have to be picking, um, the spots when you will be using one and you will not be using the other. So, for example, um, again, if you, you're using it for mean reversion, you can certainly use an unb, sorry, a range-bound indicator. But what happens if it's the market starts rallying and you want to be gauging momentum? Or the, so if you take an unbound indicator, you cannot use its absolute values in any case, it won't make sense, um, because they're not comparable across time and they're certainly not comparable across securities. So you, you're basically bouncing back and forth between problems and advantages. Um, again, I'm not saying that they're useless, it's just that you, they have limited uses and they have pros and cons. But by by using them actively, as we can, you know, later see how we solve this, you don't have to worry about any of these problems. You can have the best of two worlds in the sense of having a, a normalized scaling like the RSI, an unbound indicator like the MACD. So you can have, so you don't have to worry about picking and choosing in which cases you will be using. You just use one indicator under all circumstances.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery. Because as, as I mentioned, I started this work in 2014, around 2015, I made the discovery. Now, because the indicator itself had been normalized, you could do it. It opened up to huge number of pattern recognition opportunities that were not previously able to do. Um, so again, that was another opportunity to, to use the, um, some trading setups that were not be, you would not be able to do with the MACD.
Right. Okay. So having the best of both worlds sounds like a fantastic thing. And you just, uh, you just alluded to, um, or you told us why you were going to do that. So how did you, once you decided, I think you said back in 2014, you decided you were going to try and build something that does both, what did you do? How did you get to where you are now?
Uh, well, the idea for me was, okay, so how can I have both worlds? One idea was I can, uh, unbound and normalized, uh, range-bound indicator, which I haven't figured out how that would be because it's, it's on a percentage basis, so you cannot go over 100%. So the other idea, the other option, basically I had was to take an unbound indicator and try to put it on a normalized scale. And that certainly seemed more feasible. So the first option I had, uh, was to put it on a percentage basis. So took the MACD, which is a normal, an unbound indicator, and, uh, divided by the, um, price. I mean, divided basically divided by the longer-term moving average. So you put the short minus term minus long term divided by the long term, multiplied by 100. So you can put it in percentage terms. Um, of course, that wasn't my invention because when I looked up in the bibliography and on the internet, I was called, I think some people refer it as the PO percent price oscillator, others call it MACD percent. Um, so in any case, basically I, I looked at that, I said, okay, it makes sense. Basically, have the MACD percent. So instead of having points, you had percentage, like 5%, 6%. So said, okay, there, voila, you've solved the problem. It's percentage, and it should solve everything because, you know, we sometime in life, we, uh, compare everything by percentages and we say that's 5% an increase or a decrease. So it makes sense. But when I started, uh, so it did work normalizing across time. When I started, uh, applying this in particular markets, so I had a lot of numbers in the in the paper. So for example, I said, um, I plotted the P or MACD percent, however you want to call it, on the S&P 500, trying to find where the 95% of the values reside. Take the 5% of the values as outliers. And if memory serves, I think I found that around 2% and minus 2% is overbought, oversold for the S&P 500. So in my paper, what I did, I took three completely different asset classes, S&P 500, natural gas, and the bond, a fixed income market. So what I found is the PO, um, works for the S&P 500. So above 2% is overbought, above two, minus 2% is oversold. So said, there you go. That wasn't easy. Uh, of course, I, so that someone else had discovered it, but still, I mean, I didn't care about who discovered it as long as I could make it work. So what I did is I tried to apply the PO on a different market and try to find overbought and oversold levels. And when I tried on the fixed income market, like German bonds, I found that 2% and minus 2%, which I had for the S&P 500, were completely different. I mean, when, again, when I did statistical, um, when I tried to find again, 95% of the values when they reside, um, although to tell you the truth, although the PO had been discovered, no one had plotted overbought or oversold levels on it. So that was like, kind of like my contribution to that. I found that it was the 95% of the values for the bond were different levels. So it was at, um, minus 0.7 and 0.7, sorry, it was 0.7 to the upside and minus 0.7 to the downside. So basically, the German bond had different overbought and oversold levels. It did work for that market, but it was different for, uh, than S&P 500. Then I took natural gas, a wildly oscillating market, and I found that the overbought and oversold levels were seven and minus seven percent. Which meant that the indicator was not comparable across securities. So it could not be normalized. So yes, you can have overbought and oversold levels, but you cannot have the same ones across markets. So that really didn't solve my problem. And that means that you cannot use it for backtesting. You cannot, um, you know, create a normalized scaling across securities. I mean, if you, you know, be, if you've got a, a universe of say, the S&P 500 constituent stocks, you have to create 500 versions of the overbought and oversold levels.
Right. Okay. So then what was the solution? If it, if normalizing by price wasn't, um, you know, only resolve some of the issues that you mentioned there?
Well, the problem, the problem there was that, um, I didn't know what the problem was. And I said, okay, this is percentages, so we're comparing likes with likes. At 2% is a 2% is a 2%. I mean, why should it be different in different markets? And then I suddenly got the idea that, wait, wait, wait, wait. It's 2% is not 2%. It's because a 2% in S&P 500 is much different than S&P 2% on the bond. I mean, for a, for example, to put it in today's markets and, uh, uh, for crypto, for Bitcoin, 5% is just because some trader in Chicago sneezed. For a, for a, for a for a FX market like a Sterling, 5% day, that would be huge. That would be like a big event like Brexit. So I thought, okay, that's the, that's the difference. Let's try to normalize for volatility. They have different volatility structures. So instead of normalizing my price, I normalized by volatility. And then again, I went through the same exercise to find overbought or oversold levels. And voila, I suddenly discovered that they had the same overbought and oversold levels. And then I did it for these three markets, then I did it for stocks. I basically, I tried to put it as much. And that, that happened. Basically, normalized by volatility was the, um, was the answer. And, uh, that, so I was quite happy about that. Great overbought and oversold levels. What I called the range rules. And I used the term coined by, uh, my good friend, uh, Dr. RSI, the late Cardwell, and Andrew Cardwell, Andy, who passed away a few years ago. So I used his term range rules to create various levels, seven ranges actually for the MACD V. But, uh, that was, um, the first, you know, part of the discovery
lot, but let's keep things simple on a technical B. So if you've got, for example, um, a signal generating mechanism, a bar pattern, because I also use bar patterns, basically signal generating mechanisms, as far as I'm concerned, can come from two categories: either indicators or bar patterns. So, for example, a double inside the range breakout, that's a bar pattern. It's got nothing to do with indicators. You can use it as a regime filter. Either trend, obviously, buy signals will work better if they're in an uptrend, but within that uptrend, can you partition? Can you find other regimes? I mean, other sections in the data when it works even better? Will it work better as, um, when the MACD is above 50? When it's in a very ranging environment? When is it rebounding again? You got strong upside momentum for different reasons on the range and the rebounding range, or you can see what happens where in the in the ranging range. So again, you can, you can start looking into to fine-tuning, uh, your existing patterns based on momentum. Obviously, you have fewer occurrences, but you'll find that you have, uh, better quality setups. Yeah.
Okay, all right. Well, we're almost at time, Alex. So we'll start wrapping up in a few moments, but I just wanted to ask you, um, you, I really like this volatility normalization technique. And someone in the chat mentioned that Timothy Masters also writes about this in one of his books. You, you kind of, um, you mentioned, I think you said rate of change is one of the ones that you also tested for volatile normalization. Where else do you think that this approach could be handy to, uh, you know, just how indicators work? Can you give us some clues?
Other indicators? Other indicators? Well, actually, I'll start publishing again these on, uh, on on TradingView. Uh, well, basically, you can, you can use volatility normalization in any kind of, uh, unbound or ABS, sorry, um, absolute price indicator. So you can use it on rate of change, the Coppock curve. There are, there are many indicators which, again, I'll be publishing for free on the, uh, on the TradingView because right now, the, um, the MACDV is is being programmed into TradingView, MetaStock, Optima. Um, a lot of people have done it in AmiBroker, in its TradeStation, um, Bloomberg. So along these platforms, I will be putting these indicators as well. So to come back to your question, anything that is, um, as an absolute price, an absolute price indicator, you can, you can use it there. Yeah.
Okay, well, maybe that's a good point to, um, say, uh, I know you've got a slide here. Um, let me put this one up on the screen. So how can people get in touch with you?
Absolutely. So, yeah, if, if, uh, well, basically, if anyone is interested in knowing more about this, first of all, I can send them the paper itself. Um, it's a, it's an excellent, uh, read for people that suffer from insomnia. If anyone has a problem to sleep, by reading the paper, that should solve that. Um, also, the, the MACDV indicators, both the MACDV and the histogram itself, are right now being programmed. It should be published very shortly in TradingView. It's already in StockCharts, uh, in MetaStock, in MT4, and Optima. You can give you the code if you got it on Bloomberg Terminals and AmiBroker. Uh, it will be across all technical analysis platforms. Um, and also, I did recently, uh, I mean, I've done it many times, but very recently, I did it on, um, for the IFA, International Federation of Technical Analysts. I did a webinar for them. So send me an email, I'll send you the webinar because we will go into a bit more detail. So send me an email if you got ideas. I would really, really like to exchange ideas if someone can even, uh, improve the work I have done, um, or just to say hi. I would be more than happy to be by people. Yeah.
Okay, I watched that webinar a couple days ago during my research. It was, it was really good. So you go into more detail some of the things we kind of just brushed through today because of time constraints. So if people want to know more about that one, I'll actually put the link as well. I mean, you can contact Alex and I'll put the link in the description below the YouTube video. And to tell you the truth, the webinar I did, that particular one is quite introductory. So I have, I'll have another one which will be a lot more advanced because to again, that's just, just a closing comment. When I created the indicator, 200, I think, yeah, I found the question, 2015 up until it was published in 2022, I, I was doing a lot of work. I had no intention to make it publicly available. I mean, you know, you know, just write any paper on it. So I was working, I did a lot of work on it. So when 2022 came and I had to, I said, okay, let's make it widely available and I'll just write a paper to make it a bit more formal rather than just make a blog post. And but by that time, I had done a lot of work and obviously for the needs of the paper, I only published a certain small amount of information. Like could, so there's a lot of more, uh, meat on the bone, if you will, rather than that, just on the paper. I will be following that with more advanced webinars.
Okay, excellent. Well, thank you very much for your time today, Alex. It's been really, uh, great speaking to you and thanks for sharing your research. And, uh, I can't believe this, you published this two years ago and I've only just found out about it. So I need to get better on my research, I think. But, uh, yeah, thank you for very much for explaining this today. And, and, uh, so before we go, any closing thoughts?
Absolutely none. I'm, I'm glad to be part, part of the, uh, Better System Trader. People have, you have invited. It's a very, very good post. You do a lot of good work, uh, getting people to share their work and getting to for for us to get to know knowledge that we actually wouldn't have been able to access. So thank you so much and do please keep up the good work and continue with the podcast. Uh, a lot of people, I'm sure they get a lot of value from it.
Thank you very much, Alex. That, that means a lot to me. That's, you know, the idea behind the podcast is just to share knowledge and, and from there's so much good stuff going on in the, in the technical analysis and algorithmic Quant based communities that, uh, I think it's, it's really valuable to just speak to people, even if it's not the way you normally trade or, you know, there's always little things that you can pick up. And I really like my, I guess my big lesson from today from you is volatility normalization. So I'm going to go and do a lot of research on that. And we're getting some thanks in the chat as well. Let me just put a couple up, couple of these up on the screen and then we'll, uh, we'll finish up and you can go to bed, Alex, and I can have another coffee for for breakfast. So keep said, looking forward to more advanced webinars. And then Ohip says, thank you as well. Jay said, thanks. And, uh, here we go, one, one last one. I'm looking into M. Looks interesting. Thanks for sharing. So thanks again, Alex, and thank you everyone for joining us today. If you enjoyed the video, don't forget to hit like and subscribe, and you'll be notified of any future content that we release. So thanks again, Alex, and thank you so much.
Thank you. A big thank you to everyone. And again, uh, please do contact me if you got any questions. I'll be more than happy to share my work.
All right, excellent. Happy trading. You. Thank you. So.
[Music]