Transcription
[Music] February of 2018. I mean, it sticks in my mind. I had a whole bunch of stock index strategies and they were all different kind of strategies. You look at the historical uh performance and look at the correlations and and you'd conclude, oh, these are not correlated at all. And therefore, you think, well, I can trade them all because they're all uncorrelated. Yeah. Historically. But what what February of 2018 made uh really put into focus for me is you have historical correlation and then you have what I'd call unknown future correlation.
Welcome to the algorithmic advantage. We're here to expand the toolkit of the quant trading community and introduce investors to the many advantages of systematic trading. Our goal is to educate and inspire as we embark on a captivating journey into the vast knowledge and experience of leading portfolio managers and other experts in the field. We hope you enjoy the show and if you do, please subscribe, leave us a review, or even buy us a coffee via the link on the algorithmicadvantage.com. We really appreciate it.
Kevin Davey, welcome to the show. Super happy to have you here. Thanks for coming.
Yeah, thanks for having me. I appreciate it. Looking forward to this.
Um, let's dive right into it. Kevin, I um I know your background is uh in engineering in aerospace. I know you were doing quality control and the like and you know before your your trading career which itself is is very long now and and and that's one of the reasons that I've got you here because you've been doing this for a while and uh have got a lot to share. But just with that background, I thought that would be an interesting leadin into um how that shaped your trading and how you started to trade coming from that kind of processor orientated and quality control. And I'm just thinking about how you know people assume that trading is easy and they'll be able to get rich quick and and and just learning about proper process and uh and and procedure and quality and and working hard at at some things. So, how how did that background that you've got, I guess, shape how you approached the markets?
Okay. Yeah, it's a good question. Um, part of it, it's actually probably two parts to that. The first part is my background is engineering. So, you know, I have a bachelor's in aerospace engineering. And that kind of background, an engineering background which a lot of traders uh seem to have. Uh what that does is that kind of background makes you feel like there's equations to be solved. There's a a right way and a wrong way and if you get it wrong, uh, you know, you're out of luck and that kind of thing. And that's actually a really bad thing for trading because there are no real right or wrong answers in trading as far as well you've got to do it per this equation uh and nothing else works. There's a lot of different things that work. So that's part of what you almost as like an engineer or a technical person you have to unlearn.
And then the other part of my career before trading was in quality assurance. And so I ran uh quality assurance and engineering for a firm that made fuel pumps for jet aircraft. So basically, if you've ever flown on any kind of Airbus or Boeing airplane, our pumps were on it. And the interesting thing about that is if our pumps would ever fail, well, guess what? the engine fails and all of a sudden you're out in engine. Not a good thing to have happen ever. So, uh, the quality had to be one designed in. And so, we had to have procedures and processes that supported the design, which we actually called rule-based design. Kind of sounds like algo trading, you know, where you you'd have rules that tell you how to design something. And then the other part of it was once you had the design, you had to be able to manufacture it and manufacture it accurately within tolerance and issues would always come up and you had to constantly either uh tend to the the machines if you were, you know, grinding parts or polishing parts. You had to do that. And you also always look for ways to improve, make it easier to make parts, you know, to bring the cost down and that kind of thing. And that's that part of it really applies to trading because what I found in my own trading. So I've been trading 30 years or so and at least the last 20 has been almost all algo trading. They didn't call it algo trading 20 years ago, but what I've found is you I don't do things the same way exactly that I did 20 years ago. Same general principles, but I've continuously improved what I'm doing. And I like to think even compared to what I've how I've developed systems three years ago, five years ago, that I'm doing things a lot better now. So it's you have to continuously improve and that's you know the quality assurance part of it really comes into play there. You have to have this mindset that hey, it's never good enough. Even if your trading's going good, you've always got to be trying to go one step further. So yeah, my background before trading really helped with those uh, you know, in that way.
Really interesting. Continue improvement. Um, I'd like to um delve into that a little later as well. Just the the processes that you have built. I think that's um something really powerful that you bring to the table talking about um, you know, having a strict process for how we improve and how we develop and design and test. Uh, so yeah, I think that's a a really important theme. Let's just um, let's just cover off how you trade as well, Kevin. Like you're a futures trader. Tell me about the kind of trading that you do, the kind of time frames uh in the time types of instruments that you trade this kind of thing. Give us an overview of what what your trading looks like.
Okay. So, um, like you mentioned, I trade futures. So, I stick to the US exchanges. So either the CME or the ICE exchange which has you know between the two of them especially with the CME they just have a huge array of different sectors you know you have currencies you have metals you have energies and so on and so I'll look to trade any of the active futures that aren't necessarily brand new so if they introduce some brand new ones I don't trade those right off the bat because I'm always looking to back test to historically test whatever I'm doing, you know, so I I don't delve into any brand new markets and and I also stay away from some of the real smaller, you know, the smaller markets, things like milk. Um, I have traded it in the past, but you know that it's that one's such low volume that it's hard to uh to really come up with strategies for that and to feel comfortable trading it. So, I like to stick with the more liquid ones, but within that, I try to have a diversified portfolio. And so the way I trade, I'll look to any uh possible way to trade, meaning intraday trading, uh short-term trading, maybe a day or two, swing trading where maybe I'm in for a couple weeks, even some of my strategies are much longer term where I might be in a trade for a month. Uh, you know, so and people might say, "Wow, you can't be in a trade for a month." But think of a market like uh coffee, at least up until the la last couple weeks. Uh, coffee was a great market to be in for probably a year. And yeah, there were some um, you know, big drops every once in a while, but for the most part, it just kept going up. And so to be a long-term trend follower in that, you would have done pretty well. So, I look at uh all different, you know, uh time frames as far as how long trades last. I look at different bar sizes. So, I don't necessarily just look at daily bars. I'll look at 60 minute bars. I'll look at some weird time bars, too, 32 minute bars or 29 minute bars, things like that. And uh as far as where I get like ideas from, I I will look anywhere and everywhere for ideas. So it could be technical indicators, things RSI, moving averages. You know, most of those don't work most of the time, but there are times where they do work, and there are situations where you can use those. Uh, candlestick patterns, most of the times those don't work, but there are times where they do. Uh, statistical, you know, you could use like uh data mining. Got to be really careful with that, but sometimes that uncovers things. Um, so there's a lot of different areas that I look at and I what it comes down to for me is if I can program it and historically test it, then it's worth at least trying. But that being said, I bet 90 well over 90% maybe well over 99% of the things I try fail, especially when you consider uh slippage in commission, which is a a true cost that gets you. And a lot of people tend to say, "Oh, I'll just worry about that later." Well, you've really got to worry about it right at the beginning because especially slippage costs can just kill a strategy and uh that's what ends up happening.
Speaking about slippage um and commissions, obviously they have more of an impact at the lower timeframe trading more so than your trades that last a month or two. Do you sort of uh subscribe to the notion that the markets are essentially fractal like and that the shorter time frames are exactly the same as the larger time frames and and the same kinds of strategies apply or that there are differences there's more noise say in the shorter term and um different strategies are required there?
Well, there is noise definitely at all levels and I think what ends up happening at the shorter time frames the smaller bar sizes that you're looking at. The small shorter trade durations, the noise can be a bigger impact because for one reason is the slippage because you still have to overcome that slippage whether you're trading once a week or once every 10 minutes. And so what what it leads me or what it's led me to do, I'd love to trade 20 times a day, uh, you know, and have a crude oil strategy that just was in, out, in, out, you know, making money all the time. The reality is I don't have any of those. Um, what I've always tested uh or found the work from testing are the longer term systems where I think what's happening is you're getting more of the signal or the trend I guess you could call it as opposed to the noise and the noise doesn't matter as much but when you go down to like one minute bars you start running into a lot more noise and it's really hard to get a good trend that lasts long enough to make up for your trading cost. And that's kind of my experience. And like I said, it isn't necessarily what I personally would like. I'd love all intraday strategies that, you know, opened in the morning and closed by the end of the day and then overnight, you know, I can sleep well and not have to worry about any positions. That would be ideal. And I think that's what most people look for. The reality is those are few and far between relative to longer term strategies. So, you know, you end up at some point you've you got to tell yourself, well, am I going to force my views on the market or am I going to try to let the market guide me and I'll just follow along? And that's kind of where I'm at is I let the market tell me what works.
Across those exchanges that you trade, the CME and the ICE, like how many markets would you say are available to you to trade in terms of liquidity and history that you're happy with? What kind of um range of diversification of markets have you got there, would you say?
Well, um I trade when I test strategies, I'm usually testing 40 to 50 future symbols and they're really in uh depending how you count it, seven sectors. So you have your eggs that would be things like coffee or not coffee, soybeans, corn, wheat. Then you have your softs, which would be like coffee, cotton, sugar, those kind of things. You have your interest rates. You know, there's usually I just look at US interest rates. So, there's really only a few and they tend to be correlated, but that's a separate group. You have currencies, uh, not forex pairs, but rather currency futures. So, that's a good one. You have metals, gold, silver, platinum, that kind of thing. you have uh energies you know the crude oil heating oil that complex and then finally you have your stock indices. So out of those 40 to 50 markets you you can get quite a wide array of different sectors and that's one thing I try to I'm really pretty uh cognizant of keeping in mind when I decide what to trade. I want to make sure that I have strategies in each one of those markets and and market sectors. That being said, uh 40 to 50 markets, but I don't necessarily trade them all. Uh, there's some markets that are really hard to come up with good algos that have worked over a long period of time. Like for example, um, let's see, uh soybean oil as an example or soybean meal. I don't think I have any strategies right now in those markets, but I do have quite a few soybean strategies. So, uh, you know, I test them all and wherever I find something that works, then I'll look to use it. But I do try to keep balance when I go to a portfolio level with all those different uh sectors because I think that's important.
Yeah, I want to ask about that, but is cocoa traded on on CME or ICE? I can't remember.
Uh, that's on the ICE. So ICE has like coffee, cocoa, sugar, cotton, and then the uh those are the big ones that I trade and uh orange juice, but uh rarely in orange juice. There's not a lot of liquidity in that. And those are the ones I look at. And then also the dollar index is actually through ICE which is kind of strange cuz that's you know technically a currency. It's a basket but and then just about everything else is on the CME or you know one of their under their control.
Is um cocoa one of those markets that maybe was harder to trade and find strategies for? And uh I'm just thinking about the recent, you know, massive move over the last year or two. Yeah. Where um suddenly longer term trend followers had their day in the sun in that contract, although for a long time before you might have had a lot of trouble finding strategies that worked on it. Um, is that something you've traded and did you have the same experience?
Um, I have traded it. I do trade it. I don't have anything right now and I probably had the exact opposite experience um sort of. So let me explain. So a lot of my cocoa strategies were probably developed three or three or more years ago and they worked great. I was trading a lot of them live. Then, you know, what, two years ago, we had this massive run up and I did have some cocoa strategies that just nailed it. Uh, you know, they were they followed the trend, but then I also had some cocoa strategies that said, "Hey, I've never seen a trend like this. Uh, I think you should go short because this trend's going to end." Mhm. And and uh a couple of my strategy cocoa strategies just got killed. Now um I don't think I was trading any of those at the time because I have a a selection method which you know if if a strategy hasn't been performing well which some of those short cocoa uh strategies wouldn't have been uh I don't trade them. But it did break quite a few cocoa strategies because what we saw in cocoa and kind of what we were even seeing now with coffee uh, you know, in the last year or so were pretty it was pretty unprecedented. So yeah, you know, and that makes it hard because you're trying to build a back test or, you know, you're building a strategy to a back test that has a lot of data, but it doesn't have that sort of price movement. And then when that happens, you know, you've got to figure your some of your strategies are going to break. Um, but I'm still developing cocoa strategies and now the ones I developed today include that kind of crazy price action, you know, that crazy movement. It'll probably never happen again, but if it does, you know, I want strategies that at least hold their own. And at the worst, at the best, you know, they would do pretty well during that kind of thing. But you know, you've been around trading long enough that you know nothing ever repeats itself.
Exactly. It's always a little bit different. And you know, that's that's one reason back to that continuous improvement. You know, I'm still building strategies. It's not like, hey, I built all my strategies. I don't have to do anything else. You know, I just have to watch them and, you know, sip a margarita and and watch the profits roll in. No, you got to constantly build new strategies because strategies do break.
Yeah, you've got to work at the game of trading and um can't rest on your laurels. For context, it's the 10th of April 2025. Um, we've just seen what like a 9 and a half% up day in the S&P. I think NASDAQ 12 or something, which is its biggest up move since about 2001. Uh, you know, off the back of coming down about 20%. I mean, the markets are wild and it doesn't it it just never works out the same way twice. You've got to have your risk management hat on, don't you? Because um we we had a big sell-off in in 2020 with the COVID crash, but this is different again. And even on the chart, sometimes it'll look the same, but the impact on your portfolio could be quite different. So, yeah, they sort of rhyme but maybe never quite repeat the markets.
Right. Right. Absolutely. That's certainly the past uh 3 4 months have been like that and uh, you know, that's why I always warn people that I talk to and work with if if you're going to make a strategy for stock indices. Yeah. It can be long biased where maybe you uh are in the market more on the long side than the short side. But don't ignore the short side. And um because what you'll find is if you do that, if you're, you know, 90% long and only 10% short, something like the last three months is just going to uh decimate you, you know, or at least it's just going to chop you up a lot. So, you want to be able to have that ability for strategies to go short, to go long, and you know, if historically they've worked out, then you should be able to weather those kind of storms.
It kind of talks to the idea of building things based on a degree of logic rather than just what you're seeing in the charts or in your back tests, I should say. Um, so you know, like that risk management framework might mean that we're we're adding those shorts in even though uh we're not exactly sure when and how they're going to help, but we know that when when the time comes, we'll need them. Mhm. And uh so adding them because of because of a logical risk framework, I suppose.
Let me ask also more a little more about that that sector diversification and and also strategy diversification but just on the sectors um and we'll get into this a bit later but if as you're rotating and rolling through new strategies building new strategies and subbing them into your portfolio say what's your sort of framework and or do you have a completely rules-based process for maintaining that diversification across markets and sectors.
Um, yeah, I do. And so what I do normally, I have, uh, over probably 200 strategies that have gone through the process that I use to test them, and some of them, so they've all passed that process, then they've run live for usually a number of years. And so what I do on a monthly basis is I will uh and most of this is automated by the way. Uh, I will have it determine which strategy I'm not going to trade all 200 in any particular month. I'm going to trade maybe 20 to 30. And what I do is I make sure that I have a certain number in each sector. Uh, and I also use position sizing to try to balance out. So, for example, I don't want to have um all my or half my margin theoretically tied up with stock index strategies and 2% with eggs and softs. Um, you know, what I try to do is balance so I'm in all those sectors. Now, it doesn't always work out that way just because futures, you know, it's it's uh, you know, Unless you're trading billions where you can be pretty good with the position sizing, uh, it's kind of hard. You go from one contract to two. Well, that that's a pretty big jump. And sometimes, uh, it's hard to get that resolution with your position sizing. But what I normally try to do is I try to balance the potential amount of margin I'm using in each of the sectors. H. Now, uh, the there's a lot of issues with that and I could go on and on about it, but one is, hey, I might have some strategies that trade are always in a trade and so they're always using some margin. There might be other strategies that maybe only trade once a week for a day and you know, how do I balance that? And so, there's a lot of little issues in there that you've got to kind of resolve. But in the end, what my end goal is is I don't want to be held hostage to any particular sector. I want to be in all different sectors to some degree or at least the potential. You know, you got to wait for signals to occur. And what I found is that really helps out with uh, you know, just diversification. So, for example, um, I know like this just this month so far, uh, I was long crude oil right when everything went down. So, I lost there. Um, but I was uh short uh 30 was it 30-year bonds, the US uh ticker, and that had been going down. So, that was making money. Now like as we speak now and now it's going back up but it kind of balanced it out and then you know I was in metals and so those were going long and short and back and forth and those really haven't done a whole lot for me this month but the point is try to balance all these things out and you end up yeah one sector's down but maybe the other sector makes up for it and you end up with a more balanced equity curve. curve, which is really the whole goal is to try to get a smooth equity curve. Yeah.
So, that's a good idea. You use or you could one could use margin as a bit of a proxy for your risk exposure to that sector. And so, if you're broadly equalizing your margin, you've you've got a sense of equalizing your exposure to those different sectors. Yeah. And I don't know if that's necessarily the best. I know a lot of people use like average true range and volatility measures, but margin's not a bad one to use. Uh, you know, part of the problem with that though is if you want to go back in time and say, well, how did this work over the last six years, if you don't have that historical margin database, uh, which a lot of retail traders don't have, uh, that would be hard to use because you don't know what the margin for wheat was two years ago, uh, and how it changed. So that kind of you got to kind of have that kind of data to be able to use it effectively.
What are your thoughts on market or that sector diversification versus strategy diversification? So having different strategy types and time frames running on one market or one sector versus spreading strategies across sectors. How do you see those two forms of diversification? Is one better than the other or?
I like both. So, uh, I think I like the diversification across different instruments or different sectors a little bit better. And I'll tell you why. Uh, this goes back to 2018, uh, February of 2018. I mean, it sticks in my mind. I had a whole bunch of stock index strategies and I they were all different kind of strategies. You look at the historical uh performance and look at the correlations and and you'd conclude, oh, these are not correlated at all. And therefore, you think, well, I can trade them all because they're all uncorrelated. Yeah, historically. But what what February of 2018 made uh really put into focus for me is you have historical correlation and then you have what I'd call unknown future correlation. And you know uh what ends up happening when you're trading is the worst possible outcome is always going to happen at some point. And so what happened to me back then, I was probably trading uh I don't know, I don't remember the exact percentage, but let's just say 35% of my portfolio, futures portfolio, was all stock indices, which is a lot more than I do now. And if you go back and look at February 2018, you won't even really notice this on a chart. That's how insignificant it was. But what happened was the market dove pretty quickly and all of a sudden all my stock indices strategies all went long right around the same time which was about halfway down this drop. And of course by the time they got to the bottom that was when a lot of them either got stopped out or reversed. And so in a very short period of time, this correlation of previously uncorrelated strategies, just in the same market uh really nailed me. And then of course the market turned around after it it beat me up, turned around and you know now you go and look at that chart and you're like how did that ever happen? But that's the danger of even multiple strategies in the same market is you can you can run all kinds of correlation. You could run all kinds of analysis to say, well, there's no way that more than four of my ES strategies will ever be long at the same time. You can do that going back 20 years, but doesn't mean it's not going to happen tomorrow. And you know, I think today in the the market where the market was up and down a little bit and then all of a sudden there was that uh tariff pause and the market just took off. Imagine if you had been short and you were getting a lot of short signals, which I think a lot of people did the last few days because the market had gone down. They very easily could have been correlated had a lot of correlated strategies that said, "Oh, be short." then the market just takes off. Uh, and that's what you got to watch out for with with that. Um, yeah. So, you know, that's why I I tend to like having multiple strategies in a bunch of different instruments, but I'm always in the back of my mind. I'm always aware of, hey, I don't want to be in a position where everything goes long or if it does, I make sure I don't trade that many strategies. So, for example, with stock indices right now, I've out of 30 strategies maybe in my portfolio, I make sure I never have more than three that are stock indices. So even if they all went long, all went short, I I feel I could handle that within the whole portfolio.
Yeah, really critical. Last time we chatted, Kevin, we talked about um just those different time frames that you do trade in the different strategies. So you're trading minute bars, but you're also trading kind of daily data, if you like. Yep. Um, how do you manage trading both of those? So, you know, we talked for example of some of the nuances that can come into it when you're looking at say if you're if you're a futures trader who's just from your data provider getting end of day data, you got to be cognizant of what that end of day closing price is. Is it settlement price? Is it the closing price? What what are some of the things to consider there when we when you're trading kind of daily bars and minute bars? And how do you manage all of that?
Okay, so most of what I do now, I've gotten away from what is a daily bar, and I'll explain why in a second. And I usually use uh 1440 minute bars, which is really a 24-hour bar. Now, the markets, even the ones that are open almost all the time, they're still only open for 23 hours. So, a 1440 minute bar just gets all the data. um, you know, really could be I think a 1380 instead of a 1440. Uh, but that gets all the the data. If you look at daily bars um it depends on your data provider how they identify the closing price. So this goes back gosh at least 20 years 30 years to when there were pit there was pit trading and they had much smaller hours. So, let's just take gold for example. Gold used to trade, I believe it was uh it was in New York, it was like 8:30 in the morning and it would close at 1:30. So, the settlement price when they figured out, hey, the, you know, the closing price for that day was determined to be between 129 p.m. and 1:30 p.m. Okay. So, you'd have a daily bar and it would reflect that price at the close. Makes a lot of sense, right? Well, then they went to the 24-hour markets. Uh, and it didn't all happen at once. It kind of expanded over time. Well, now, uh, they never changed the settlement time, at least for most markets. Maybe some they did. So, gold, the settlement price is still around 1:30 p.m. every day, even though it keeps trading till 5:00 pm. And if you don't know that, and you don't know what your data provider provides, you might be thinking, "Oh, that was the price at 5:00 p.m." when really that price was at 1:30 p.m. So, just to give you an example, I use Trade Station. Their daily bars, their closing price is their the exchange settlement price, but the 1440 minute bars, the closing price is the 5:00 p.m. the last price traded price. And um on days, well, today almost was a good example because I think the the tariff news came out like around right before 1:30. But let's just say that tariff news came out at 400 p.m. Well, you'd have a lot of markets like gold, oil, uh, interest rates that all settle before that time. So, there's closing price settlement price would be a lot different than the last price traded. And of course that is makes a huge difference sometimes on all your indicators. If you're calculating moving averages or you know whatever you're calculating breakouts you could get wildly different signals. And so you've got to kind of account for that. And I'll tell you, most people, including myself, never even realize this was an issue until you see a trade and then maybe you refresh the chart and the data changes because now it it updates the settlement price and your trade disappears and you're like, what happened? And then it leads you down that rabbit hole of, oh my goodness, this this data is all different. And uh, you know, so depending on your data provider you could take the exact same strategy. I could you could run it with one data vendor I could run it with another and we could get totally different results just because of that difference. So it's it's one of those little nuance things but uh, you know, if you're into it for a while you'll see it.
When you're trading the 1440 minute bars, do you have the opportunity to say get a closing price? Let's say it's at, you know, 300 p.m. and the market reopens at 4 and then execute say a trade that you want to get in in that hour gap. As if to say, you know, if you're trading kind of those daily bars, you've got time to process by the end of the day before the open the next day as opposed to trading with the 144 minute bar. Maybe if you're trading like daily bars like you're waiting for your data provider to give you the end of day data which is going to happen after the market close and then then you're going to process your algos and then put your orders in and so the orders are technically late. Um, can you can you um can you act quicker with those 144 minute bars?
Um, the nice thing about the 1440 minute bars and the reason I use them is because you know that closing price uh really won't change. So when the markets close that price you know your strategy will run with that price and it everything will be okay and it will give you a signal or not. With the daily bars, what you have to rely on, at least with Trade Station, and again, they use the exchange settlement price for their closing price. If you look right a second before the market closes at a daily bar, it will show the last price traded. But then sometime in that hour of downtime what's supposed to happen is uh Trade Station all the other, you know, platforms would be notified of the exchange they would get that settlement price should be pushed through to your trading platform the chart should automatically update with the settlement price on it. That's the theory. The reality is that doesn't always happen. Um, I've seen cases where the exchange doesn't provide that data until 15, 20, 30 minutes after the market opens for the next day. So, sometimes it's the exchange, sometimes it's the the data provider doesn't update their data, sometimes the charts don't refresh. There's all kinds of little issues there. And what ends up happening is sometimes you will actually get what I call a phantom trade where it'll take a trade but then if you refresh the chart with up-to-date data the trade will disappear and then you'll be like if you notice it you'll be like, oh man, now what do I do? So um, you've got to watch out for that but you know that is why daily bars are not good. But what is good about daily bars are those are the settlement prices. Okay? So everybody in the world who gets a brokerage statement, their brokerage statement's going to reflect that settlement price and and that's useful for, you know, how much equity they have, how much margin uh leeway they have. So, uh, you know, a lot of people put a lot of faith in that settlement price because that's what you're marked to at the end of the day. So, you're not marked necessarily to the last price traded. You're you're marked to the settlement prices. And when you look at it that way, you can say, "Wow, those settlement prices are important." You know, you think of like a big grain house that uh at the end of the day they see that settlement price which might be different than the last price traded, maybe not. And they might make some decisions based on that settlement price. You know, they have obviously certain risk tolerances and you know, they have to that's what they have to look at. So, there's pros and cons to each. Um, but I found for automated algo trading, I've gotten kind of gotten away from daily bars more and it it really is more of a uh just a practical functional issue. You know, I I don't want a lot of issues with trades disappearing and data not updating. So, that's why I tend to do what I do.
H you know Kevin you you run some workshops so you've been helping traders for many years uh as you're so you're you're looking at retail traders you understand the um the issues they face I guess um in a world where there are ever more quants with ever more data and processing power and machine learning and um uh tools and capabilities at their fingertips. Um, you know, do you do you see that the retail the the ability say for the smaller retailer retail trader to have an edge is uh decreased or or increased or has remained the same? Like how does how does the newer retail trader um get ahead?
Uh, that's a good question. Uh, and most retail traders probably don't get ahead, but um, I'd like to say, and because people ask me this, they're like, "Well, you've been doing this for 20 years. Isn't it easy now?" And my response is, "No, it's not. It never gets easy." And, you know, people are always perplexed by that because they're like, "Well, you've been doing it for so long, you should be so good at it. You should be, you know, making money in your sleep kind of thing." like all those YouTubers, Kevin. Yeah, exactly. And you know, the the thing is, uh, every day there's new competition and like you said, there's new tools out there and people are using those tools. The good part about that, all these new participants, is usually people come into this with uh just a twisted mindset that it's going to be easy to make money. It's going to be easy to develop strategies and you know it there's not going to be any like really tough things to do. They just got to do it. Uh, you know, I just yesterday I got an email from somebody who said I've been trading a year and a half and I'm still not profitable. What am I doing wrong? I'm like, well, you know, who knows, but you could say you've been trading 10 years and still not making money and that wouldn't surprise me necessarily. So, I think it's uh I guess it's harder than it was probably 20 years ago when you had nobody doing back tests and that kind of thing. And at the same time, it's easier to do the back tests, but uh you still fall victim to typical human thinking. You know, I remember I mentioned the beginning of being an engineer, a technical person. and you think, I've got to solve this problem. I got to solve this equation of the markets. That's not necessarily how it works. So, that makes it uh tougher for people who just don't come in with the the right mindset. So, overall, um I don't want to say it's easier because it's not. The tools are certainly better and you can do more, but now there's a ton more people trying to do the same thing you are. And you know, futures are a zero-sum game. Every if I'm long gold, every dollar I make in gold, there is somebody out there losing money. They're losing that dollar. Now, they might uh people get confused with this cuz maybe they're hedging or something and they've got money somewhere else in real gold or, you know, so but that contract, it's dollar for dollar. There is a winner and there's a loser and it might get transferred around a lot but uh, you know, it's a dog eat dog kind of world with futures where that's different than stocks and that's where I think people get messed up, you know, uh people who bought Tesla 20 years ago or whenever it first came out, I don't even know, uh they might have all made money and it doesn't necessarily mean somebody was on the losing end of that. It was because Tesla grew as a business and everybody all the owners benefited from that and futures is different. So that's it.
Yeah. Before we move on to the next stage because I want to get into your strategy development philosophy and process. Okay. Um, I just want to quickly ask you if there's a a particular style that you lean toward like a particular say marketing efficiency or behavioral pattern that um that you like to exploit the most. For example, are you more of a mean reversion trader? Are you more of a trend following trader? Do you see yourself in in do you see trading in in that light or um do you have a different outlook alto together?
Well, I think uh for the most part a lot of what I'd be uh trading would be considered trend following only because I tend to go for fewer trades with bigger average net profit per trade. So, um that leads me just to not trade as much, which is nice. I don't want to be trading 10 or 20 times a day in a particular market. I don't think there honestly there are 10 or 20 opportunities every day and that have lasted for a number of years that you can take advantage of but there might be a couple a month or a couple a week. Those are a little more likely. Um, so that being said tend to do trend following. The downside to it is if you're in a long-term trend so I mentioned coffee before. Let's just say you had been long coffee for the last 6 months. Yeah, you would have made a lot of money. But if you were to look at a chart and just say, "Hey, last April 1st, just to throw out a date, I'm going long coffee." And if you'd look at that equity curve, you would see some pretty scary drops, some big losing days, probably some big losing weeks, maybe even a losing month, but overall, you would have made money. And what that means is you have to endure a lot of pain or at least a good deal of pain to get to that end result. And I think a lot of my strategies that I end up trading do have pain to them. You know, they are not these riskless, you know, in and out where the equity curve is super smooth. I'd love it if it that was the case, but uh those either they don't exist. I've never been able to find them. The one ones I always seem to find have drawdowns in them.
It's the medicine you have to take we say. Yeah, absolutely. And it's hard uh for most especially retail traders to buy into that. You know, uh the example I always give is is pretend you are a mountain climber and you're going to be climbing that equity curve that you just created and you'll realize it's not a nice uptrend where you, you know, it's like an escalator. You're gonna be climbing and all of a sudden you're gonna fall into a creasse and then you're gonna pull yourself up. You're gonna fall again and you're gonna get tired and you know it's tough. It everybody would what they usually see they see the start point and the end point. Well, if I had been long coffee last April, I'd have this much money now. And they dismiss all the in between stuff. And that's the the super important part of it cuz if you can't make it, if you can't stomach it, and maybe that's where some of your edge comes in is being able to have the confidence in what you're doing and how you develop the strategies to actually say, well, I can withstand this because it's happened before and I think it'll, you know, normally it'll come out of that. That's that's hard. That is really hard.
Say that's um part of your edge. I've been thinking about this concept of edge recently and wondering um whether it still has the same meaning as it might have back on the days in the pit. Um, because especially for an algo trader, you know, we're doing a lot of hard work and applying a lot of good process to, you know, to get the result that we want. So, um do you ever think in terms of what your edge is?
Normally I don't. Um, and and there are strategies I have that uh seem to work, but I don't even really know why. Or worse yet, I thought something would work, but I test the reverse of it, and the reverse works. So, you know, there goes all my brilliant thinking. It's actually the opposite. So, um I tend to think of any edge that I might have might not be strategy specific, but it might be more um process specific or trader specific where uh I know what to expect. I know there'll be draw downs. I don't have unrealistic expectations and uh I'm willing to stick through the hard times and I think there is kind of an edge in that just being able to withstand it uh because there are going to be those bad times.
Yeah. Um, okay, so getting into your strategy development process and then I want to get into your robustness testing but just in the strategy development I guess if we can skim through it. Um, you know, do you have uh a an approach of or where do you fall in the spectrum of sort of simple versus complex strategies? Do you tend to try and make them as simple as possible or in pursuit of that edge you end up uh developing quite a bit of complexity?
Well um so one the first
The thing is, I'm always, uh, driven by what the data tells me. So, uh, personally, I'd probably like more complicated things, just having a technical background because I would think, "Oh, you can account for more things that happen and it might be better." But what I've found is usually it's the simple things that work better. And, you know, strategies with a couple lines of code as opposed to thousands of lines of code, those tend to work better, at least the way I develop them, and the way I test them, and the way I trade them.
Um, you know, there might be some people out there who have, you know, supercomputers that are doing all sorts of, uh, you know, millions and millions of, uh, runs to figure out what's best. That's never worked for me. It's always trying to keep it simple. And unfortunately, what that means, going back to the drawdowns, that means you're going to have to withstand more drawdowns because you're not always a simple strategy. Let's just take a breakout. Simple 20-20 bar breakout, you go long if it's a high breakout, if it goes, uh, a low breakout, you go short. Uh, that can actually, if you do it right, that can actually make money in, in different markets, but it's going to have a lot of pain along the way. There's going to be a lot of breakouts that fail. Uh, and so then you say, "Okay, well, that's where you, you think, 'Oh, well, I'll just add a rule to filter out these breakouts.'" And then you start running down this rabbit hole of creating the perfect backtest. M, and that's where people get messed up. So, um, I tend to keep things simpler and live with the nastiness sometimes of what I've tested.
How much, um, how do you balance the need for, say, a strategy to have a sound theoretical basis, um, or a causal basis versus, um, just accepting the empirical results that works, something that you've, that the machine has thrown out?
Well, I try to avoid ones that are just like randomly generated strategies. Uh, but at the same time, uh, I like, personally, I like to come up with an idea, but I'll test it. I'll also test the reverse. I might test a couple different versions of it. Uh, but I, I like, just from a, I guess, a confidence point of view, to sort of have an understanding of what's going on and sort of be able to rationalize why it works. But at the same time, I realize there's a lot of things that I don't understand that probably work, and I have no idea why. Uh, you know, a good, for anybody who's, who's kind of on the fence about that, read the book, uh, *The Man Who Beat the Market* about Jim Simons at Renaissance Technologies. You'll see there's numerous examples in that book of where they say, "Hey, this algo worked and we have no idea why, and it still works, you know, or it worked for years and nobody could figure out a reason, but it worked and it passed all our tests." So, we traded it, and guess what? We made money.
So, um, you know, you don't necessarily need something to make sense. Uh, as long as you put the same kind of test to it, whether you came up with the idea, you found it on the internet, or you just randomly picked one. Um, you know, there's, I think all could work, but personally, I like ones that at least I can sort of understand. I don't like ones where I just throw stuff. And, you know, I know there's software out there where you can literally just go down a list of, yeah, do RSI, do ADX, do stochastics, do percent R, do, do this, do that, do that, and mix them all up and come up with a strategy. Uh, you know, that, I'm not a big fan of that personally.
So, um, just finally on the strategy build stage. So you said, you know, you could test out an idea before really trying to validate it, just in that initial strategy build process? Because I know we'll talk about walk-forward analysis because you do that, and obviously we've just had a couple of interviews with, um, with Bob Parardo where we got into this in some depth. But when you're just doing that initial build, do you say, cordon off a little bit of data at the beginning of a period or just take a certain bit of data to run through that build and leave a bunch of data out-of-sample? And I'm asking if you, I guess, h- how do you spend time kind of building the initial idea before going on to really validate that it is going to be robust?
Mhm. So there's really two different paths I take. And so sometimes I take one, sometimes I take the other because one of, and one of the reasons for that is I like to not think that I'm always doing things right. And so I like to keep myself open to other options. So one thing I do is, yeah, I'll, I'll play around with the strategy. Maybe I'll take an idea, but it's not really well-defined, and I don't know what I should put in it. Um, I will test that, and I might test it quite a bit, but I'll only use a little piece of d- of historical data. I don't do all the historical data. Uh, and there are people out there who do that where they'll take all 20 years of their historical data and they'll say, "I'm going to try a breakout." And they look at, you know, look at the results and they say, "Okay, now I'm going to add a filter to this and test over the same data." And they keep testing over all their data, and maybe they have success with it, but I've never had success with that. It always leads to overfitting.
So, the one route I take is just a little chunk of data. And I can play around with things. I can change variable ranges, you know, whatever I want to do. And I know I'm not going to be curve-fitting because ultimately I'm going to be testing on the big data set. Okay, so that's one way. The other way is I just say, "Hey, I think I have a decent idea. I'm just going to test it on all the data." And so I'll literally go from idea to build the strategy. Here's my strategy. I know all my inputs, whatever optimization I want to do. And I'll immediately go to the walk-forward optimization, which is when you use all the data. And the key with it is it's a one-shot test. So it either passes or it doesn't. You can't go back and change it. You know, you can't get the walk-forward results and say, "Ooh, you know, I like it except in 2022, had that big drawdown. So, what if I put a filter in for right there?" And then run the walk-forward again? Because the idea with walk-forward or any out-of-sample testing, that's what walk-forward is when you do it right, is you do it once, and then it's truly, you know, unseen data that your strategy never saw it. But if you run the walk-forward testing correctly and you, you do a hundred versions of the strategy, it's not really out-of-sample anymore. It's just, you think you did out-of-sample, but you really just ran an optimization yourself. Yeah.
So, we're starting to talk about your robustness testing and validation process. So, you've mentioned Walk Forward. Can you maybe back up a bit and just give us the, the 30,000-foot view of your process? Uh, you know, the key parts of your validation process and, and where Walk Forward fits in that.
Okay. So, the, the first thing I start out with is just knowing upfront what I want as far as goals and objectives. And most people don't. They just say, "I just want a strategy that makes a lot of money." But you really got to have those goals upfront because that's your cue to know when you have something good enough. You know, if you meet your goals, then it's time to stop it. And again, it, it goes back to what most people want to do is they want to perfect what they're doing. And yeah, that works in a lot of different aspects of life, but in building algos, it doesn't, at least in my experience. You want to make it good enough, but not too good. So that's where I start. Then I come up with the idea. I do some of my preliminary limited testing, small data set, and then when I'm ready to go, I'll do the walk-forward testing. And I take the results of the walk-forward test and then I'll do some, uh, what's called Monte Carlo simulation. So, I'll take my trade results, mix them up. Uh, you know, it's a process to give you some probabilities of, hey, what's the probability that I'll get wiped out if I trade this strategy with this amount of money for a year? What's my return to drawdown? U, that's to me the most important metric there is. It's not net profit. It's not max drawdown. It's the profit to drawdown. It's the combination of those two. Um, so that becomes like one of the keys. And so I have certain thresholds and I don't necessarily have a robustness test for it, but what it, the limits I use or the, you know, the boundaries I use because I say, "Well, if, hey, the return to drawdown's below a certain point, it's not worth trading." But if it's too good, well, you got to worry about that too, you know. And what I always go back to is thinking about, if you faked a good trading system, how would you discover it? How would you determine it? One way is by if the results look too good to be true. Those things fall apart. And, um, so just for doing this years and years, I have those limits, and that becomes a pretty good test of predicting future performance.
Um, but the big one of the biggest things is seeing how that strategy performs after you're done developing. So you, you stop development, you run your walk-forward test, you do your Monte Carlo simulation, you say, "Yeah, this is in a good range, but I don't know. Maybe I overfit, maybe I curve-fit, maybe I did something wrong." So, you let the strategy run live, not necessarily with real money, although you can. Uh, that kind of defeats the purpose, but you want to make sure the strategy doesn't fall apart. And how long would you, would you give it? I usually do anywhere from like six months to a year. And it's a long time. Very patient. Well, that's, uh, I especially know that when I first started doing it, I was like, "Okay, well, I just developed a strategy. Oh, now I got to wait nine months. What?" You know, what am I going to do in the meantime? But the nice thing about that is once you get rolling with developing strategies, you're always going to have some new ones coming up that were nine months ago and have been nicely validated in a live market. I can't think of anything better.
Well, yeah, it, it doesn't necessarily mean, you know, I've had strategies that that tested great in walk-forward testing, that looked great during this nine-month period, and then I start trading them live, and they fell apart. Um, actually, I had one last week that was, uh, an intraday system for ES, and, uh, literally the first day I started trading it live with real money, uh, it, I lost per contract $6,000. I was like, "Yeah, well, it's been a rough time." Yeah. I'm like, "How did that happen?" You know, it has this great-looking equity curve, and as soon as I went with real money, so, you know, if I had waited one more day, wouldn't have happened. But so it's no guarantee, but there's definitely, from everything I've looked at, a correlation between, hey, you have a good walk-forward backtest, and then you have a good nine-month live test, for lack of a better term, that that combination tends to lead to strategies that perform live better. Doesn't guarantee it by any means, but it kind of puts the odds in your favor a little bit.
And so, yeah, they, you know, you talk about robustness tests and what a lot of people do, I think they make the mistake of putting their robustness tests on their historical testing. And, you know, they'll say, "Well, yeah, my backtest looks great, and I ran all these statistical tests on my backtest, and everything looks great, so therefore it's good." And my counter to that has always been thinking, well, what if that backtest was fake? You know what? If, what if, um, you and I were working together, and you decided to give me a fake backtest because you wanted to impress me, or I wanted to impress you, and I gave you a fake backtest, and you would look at it, and it looks great, and you run all these robustness tests on that backtest, everything looks great, and we both agree, yeah, that should be a good strategy, and then we we go live with it, and it falls apart. That live test, it's hard to, you know, it's hard to really cheat that. Um, and that's one reason, uh, you know, students I work with who take my course, I, uh, I always survey them of, you know, what do you think of this step? How does this step work? And this step, which I call incubation, because what you're doing is you're incubating the strategy. You're just watching it. You're not trading with real money. Uh, every time I ask people that, usually about 85 to 90% of people who answer say that step has saved me money by either avoiding bad strategies or, um, you know, waiting and then only trading the good strategies. It's, it's basically, uh, really helped them out. So that turns out of all the robustness tests I think you can do, I think that's the best one. But it's also really, like you mentioned, it's really hard to develop a strategy today and wait nine months before you go live with it, you know. Yeah, most people cannot do it. And, and that's, but again, I think that's what makes it so useful is, hey, if, if nobody can do it, that means most people aren't doing it, and they're going live with all their backtested garbage, uh, and then they lose.
We should remind you that the conversations on this show are informal and for entertainment purposes only. Certainly, any general advice you may hear is obviously not specific to your needs, goals, or objectives. So, nothing discussed on the show should be considered as investment advice. If you want that, you'll need to actually do your own research and speak with your financial advisor. Remember, trading can be extremely risky, and past performance is not necessarily indicative of future returns. If you enjoyed the show, please subscribe or leave us a review. And if you have any questions or feedback, we'd love to hear from you. Bye for now.
[Music]