Transcription
This is undeniable proof that you should stop optimizing for the highest possible win percentage. It can get very dangerous. So, let me explain more in this episode.
Hi, my name is Thomas, also known as the Mr. Breakout. I'm a breakout trading specialist and an internationally recognized trading author. I've traveled the world trading breakout strategies and on this channel I will teach you my own way of finding the right trading opportunities and making real money with breakout trading.
Hello. Hello. Welcome breakout trading fans, breakout trading specialists, breakout trading masters. Today we're going to talk about a very important topic. So, you know, a lot of especially beginning or inexperienced traders, what they do, they they optimize their strategies for the highest win percentage possible. All they want to be is to be right, right? Like they want to be right on every trade. They want to have maximum of winning trades and then they're tweaking their conditions and adding filters like crazy to achieve the highest win percentage possible. And I'm going to show you why you have to stop this practice if this applies to you and if not, you will understand very clearly what is the danger and what is a much better function to optimize for. Okay.
So let me explain what we have here and how we're going to do this. So right here I have imported, I have imported about two and a half thousand strategies. So what we're looking at, we have two and a half thousand strategies and these strategies have so-called in-sample data and out-of-sample data. Now, if you still do not understand what is in-sample data and out-of-sample data, you might need to Google it. But very quickly, in-sample data are so-called seen data or the data we build our strategy on. So these are the data we know in advance, we work with seen data which we use to build the strategy, whereas out-of-sample data are so-called unseen data. So unseen data are the data which do validate your strategy on unseen, quote unquote, future unseen data once your strategy is created. So it's a very important thing because then we can see, uh, what would be the result of our strategies built on in-sample data, what would be the the potential, uh, outcome on in true life trading, right? Like in out-of-sample trading. So it's kind of like a simulation, what would happen once I take these strategies and run them live, how well would they do and how many of them would do well? So, uh, that's that's the in-sample out-of-sample explanation.
Now, as I said, we're we're working with a pretty big sample size, which is two and a half thousand strategies. These are breakout strategies. Doesn't matter the market, time frame, uh, all the details, even in-sample out-of-sample split does not matter for this, uh, particular explanation at all. These are nuances and it will not change the outcome. By the way, the strategies, if you're asking how to build two and a half thousand strategies, all of them were built with this formula called the Mr. Breakouts model. It is the model that I use for all the strategies. If you're interested in more information, definitely check this book, thebreakouttradingrevolution.com where I describe the entire formula in details with a lot of examples how I use this formula to travel the world, uh, lot of other case studies, very, very good and full comprehensive resource. Or if you want something very quick to start with, there's a brief overview of the formula on the link below. So you can go on the link below and grab this, uh, this, let's say, um, simplified version of the formula with a quick explanation, something to start with.
All right, now let's get back to our analysis. So we have in-sample and out-of-sample data and what we have here is we have in-sample win percentage, win, win, win, win, win, win, win, win, win, win, win, win, win, win, win, win percentage, percent profitable or win percentage for in-sample and out-of-sample and we have this super important chart. So basically, this important chart tells us if we take an in-sample strategy and each dot represents one strategy. If we take it and then take a specific, uh, in-sample win percentage, what would be the win percentage in out-of-sample, which is this vertical axis. Okay. So, for example, here we can see the in-sample, uh, win percentage, or this one would be about 60% this strategy, but only about 53% in out-of-sample. Okay.
Now, what's really important about this chart? So before we get to the chart itself, the most important metric here is correlation. So the correlation really tells us the dependency between the in-sample win percentage and out-of-sample win percentage. And, uh, no correlation means no dependency, that means completely random outcome is zero, and high dependency, the best it can get is one. And here we have 0.11, that means 11%, uh, correlation. So 11% correlation itself is already telling us there's nothing to find, like there's no dependency. There's no relationship between in-sample win percentage and the out-of-sample percentage, uh, win percentage. We could easily wrap it over here and say, "Hey, it it's a waste of time, right? Like why should you optimize for win percentage?" Because win percentage in in-sample does not guarantee or anyhow, um, impact the output win percentage in live trading, like at all. The correlation 11% not at all. But of course, we're professionals. We want to go deeper, uh, than that. We need a deeper explanation. So, uh, we'll we'll we'll do more than that. Okay.
And now, a very quick shout out to our sponsor, Breakout OS. In the new era of AI, everything you know about building trading strategies has changed completely. All the generalist principles no longer work. And that's why Breakout OS is a complete game-changer. It is a highly specialized no-code platform built for the AI revolution. It's the choice of true breakout trading professionals, from leading critical traders to hedge fund managers. This will fundamentally transform your workflow and change your trading forever. See the paradigm shift for yourself and sign up for a completely free walkthrough at www.breakos.com.
Now, so that's the first thing to check and you already have the answer here. Now, the second thing to check is this, um, is this red line. So if there would be dependency, we and really strong dependency, we would see a line more like from this left down corner to this up right corner, and that that would s that would mean that, uh, with each higher in-sample, uh, win percentage, we're also getting higher out-of-sample percentage, right? Which is what what we assume or what this is what the traders optimizing for high win percentage assume. They assume, "Hey, I'll make this strategy on my in-sample data 80% win percentage and I'll be making money on every, every, every trade or at least eight out of 10." Well, it's not going to happen because this alone already is telling us this is pretty flat, that again, uh, high win percentage in in-sample says nothing, literally nothing about high win percentage in out-of-sample. Anyway, uh, so that that's this picture.
Now, what we have here also is this quite important thermometer. And this thermometer basically, uh, is taking the mean win percentage on out-of-sample. And if we, if we narrow down and we'll, I'll show you, we'll do it with this slider. Uh, if we narrow down the win percentage, how the mean out-of-sample would mean the average, the average win percentage out-of-sample, how would it move? Now, what we can see here is that the, uh, the mean out-of-sample on average is higher than, uh, mean in-sample. Okay, 56 and a half percent versus 59. So now you could be saying, "Hey Thomas, but isn't this the proof, uh, that that, um, I can get even higher win percentage in out-of-sample?" Way, not so fast. It's more of a deception. It's it's not really how it works. Uh, the most important again is the correlation, this line which says which is pretty flat and, uh, right? So like everything we're looking at is already telling us, "Hey, if you optimize for a win percentage, it does, it's got no impact on true out-of-sample data and live trading." Okay, there's like it's a waste of time. Anyway, it's it's actually even worse and I'm going to explain how it's even worse.
So, these unexperienced or amateur traders, what they do, they not even, they completely ignore this reality. They don't even know about this reality. And by the way, uh, if you, that's why you have this, uh, channel. So don't forget to subscribe so I can give you way many more, uh, important information and and lessons like this. But anyway, what they do, they say, "Hey, I just developed a strategy in-sample data. It's showing 60% prof, uh, 60% win percentage. Six, six out of 10, uh, trades are winning trades. But I'm so smart that I'm going to add five more filters and I'm going to optimize all these other input parameters and I'm going to make it 80%." Right? Of course, unexperienced traders, what they do, they overfit, they overoptimize, they launch it live, and then their live strategy is like, blah, right? Like down, losing money, and they're wondering, "Well, what did I, trading is not working, blah, blah, blah." Anyway, I want to show you if we simulate their behavior, what actually happened. It's quite mind-boggling. Okay.
So, if we go back to this example now, we can simulate, uh, the behavior of those amateur, unexperienced, naive traders. And we can simulate it by, uh, grabbing this slicer or slider, whatever you want, and then pushing towards higher and higher, um, win percentage. That's exactly what these guys do. Okay? And we start pushing and pushing, and now look what happens. Okay? So we say, "Oh, these guys say, 'Oh, 50%, I'm a Forex genius. I want 100%.'" So they start pushing and pushing higher and higher. And now look what happens. The higher we push, okay, the lower the out-of-sample mean is. So the more we're trying to optimize for even higher win percentage, the lower the average win percentage is in out-of-sample. The mean, the average out-of-sample. Like, get this, guys. The more we're trying to push win percentage on our in-sample data, the worse it will get in live trading. So if you're pushing to 70, 80%, then you should not be surprised like your equity is down and losing like crazy because that's what that's what's going on. That's like the more the more I'm pushing. Have a look, like even this red line is now going down. We're way below the mean. So everything is going against you. Like the more you're pushing, the higher win percentage in most cases than not, if you're pushing like this, you're overfitting, you're overoptimizing. And I'm not, uh, I'm not now talking about experienced traders who can distinguish these nuances and they think about the whole thing a little bit differently, right? Like I, I'm mainly talking about these naive, beginning meta traders, geniuses in forex, and guys like these who do not understand these basic principles. Very, very important. Okay.
So, do not do this. Do not push. Like you see, you're pushing higher and higher, and you're getting worse and worse results. That's how unexperienced traders get very quickly disappointed. That's the reason. All right.
Now, you can ask, "Thomas, so what should I do? Right? Like, what should be, what should I optimize for?" And there's not a single answer because there are a lot of, uh, different ways and metrics to optimize for. And, uh, actually, I'm going to, I'm going to cover more of them also on specific markets or specific strategies, um, sometimes in the future. So please don't forget to subscribe and make sure that you will not miss it because this show is, uh, every Wednesday, and every Wednesday I'm delivering new studies and strategies and tips like this. And by the way, if you want me to go through some specific metric, drop a comment below and let me know which one you want me to cover. Okay.
But one, and that's that's quite awesome. One which you definitely can start with is the old, obsolete, evergreen net profit, right? Net profit sounds too basic, too, right? Like everybody talks about net profit. You know what, guys? It works. Like it's a good metric to change. So let's reset this and let's now go for the net profit. Okay, so we change for the net profit, right? And now let's have a look at the picture again. Okay. So first of all, correlation, all of a sudden, almost 63%. 63%? That's a lot. That's one of the best correlations you will ever, ever see in trading, period. That's very high. So you can see there's a very strong dependency between in-sample net profit and out-of-sample net profit. Plus, you see the the red line is already pretty much what we want to see, almost coming from one corner to another. You see the slope, like going up. That's exactly what we see. So there's a very obvious dependency. The higher the in-sample net profit, the higher the out-of-sample, uh, profit. Okay. So that again, here we could stop and say, "Hey, guy, that's it. You see, like net profit, good one." Anyway, now let's see something even more awesome.
Now, if we push higher net profit, look what happens. Have a look. You see, do you see the thermometer? Okay, it's increasing. It's increasing and increasing. So the higher the better, the the net profit, uh, on the in-sample, the better on average and out-of-sample. So, uh, it doesn't even go to the left. It goes just right to the right side. You see, uh, and at some point, it kind of stalls or struggles. But overall, like we can see that if you, let's have a, let's say you have 500 breakout trading strategies and you do not, you don't know which one to pick for live trading. You need just 10 of them. Well, prioritize based on, uh, net. Just if you only prioritize on in-sample net profit, you're already not doing anything wrong. Like you're doing the right thing. Opposite of, uh, win percentage, right? If you would be picking with high win percentage, you'll be like, blah, down. If you're using this one, uh, net profit, you're pretty much giving yourself better and better chances that the strategies you pick will do pretty well in the live trading. Okay.
Again, if you don't know how to create 500 strategies, don't forget to, um, either download the material completely free on the link below. You definitely can create 500 strategies. You can create even a thousand strategies, 2,000 strategies on any market in the world. Uh, read this book, uh, which goes really, really into the depth. And now, uh, let's get back. Okay. So we can see that net profit does impact quite a lot the numbers, right from the beginning, talk, uh, in our advantage, quite a lot. So that's what we, uh, would pick. Okay.
Now, does it mean you're still allowed to do the crazy overfitting and stacking filters on top of filters and all this fine-tuning? No. No. Not at all. Okay, guys. So, uh, it's, you still need to do this reasonably. You need to be clever. You need to do it well. Again, read my book to find more information. Uh, and it's easy to, you still can overfit and overoptimize heavily with net profit, uh, as a, as a metric to chase, but you're already doing something right. You're already doing something much better and smarter and wiser and statistically sound and backed by real data than just chasing high win percentage. So again, never ever chase win percentage.
Let me know which metrics you would like me to explore or on which markets in the future. Let me know, uh, in the comments below. Also, like this video if this is the kind of stuff you like and you want me to create more about. Subscribe again. Next episode, it's coming already next Wednesday, and I'll give you another tip on how to make better strategies, how to build breakout strategies, how to be smart about trading, and how to make real money. And that's really all. So, see you in the next one. Set your spirit free.