Algorithmic crypto trading

Transcript

Andrew Swanscott 0:00: Welcome to Better System Trader Live, the show for systematic and algorithmic traders. Welcome to this show, we're going be talking about something that we don't often get to talk about. It's a topic I think maybe in this industry, there isn't a lot of, or that I know of a lot of legitimate systematic or algorithmic traders.

But today, the guy I found for you is, I think he's amazing actually. He's been on the trading panel the last couple of weeks, so you may recognise him, but he is an algorithmic trader, crypto specialist. Welcome to the show Pavel Kýček.

Pavel Kýček 0:36: Thank you, Andrew. Hi, thank you for having me. I have to say that I love your podcast, so

Andrew Swanscott 0:43: for me, it is really honored to be here. Thank you. Excellent.

Well, it's great to have you on BST Live. As I mentioned, you've been on the trading panel a couple of times and you've always got a lot of good insights and comments to share there. So I'm really looking forward to having this conversation with you today.

And as I mentioned, our topic is going to be algorithmic crypto trading. So I discovered you actually on Twitter, I don't know, maybe a couple of months ago, right? And I think it was something about robustness.

I was like, don't often see people in the crypto space talking about robustness, drawdown portfolios. I was like, this guy's good. So we had some discussions and I've looked into some of your stuff and we're going to dig into some of that today.

So I'm really looking forward to our chat. But first, how about a little bit of background on yourself? Can you tell us a little bit of your trading background so we can have some context around our discussion today?

Pavel Kýček 1:43: Yeah, of course. Well, the topic robustness, by the way, is one of my favorite ones. So I'm really looking forward to it.

So my background, I have to say I started as typical discretionary trader, so I thought that I will outsmart I will outsmart markets by looking at the charts as often as possible to get a feeling of the charts — probably all of us know it. So these were my starts in, let's say 2007, 2008. Then I started trading like semi discretionary half automatically or half systematically.

I used to be trading like mean reversion strategies on E mini equities and the level two data. So this was the first time after, let's say five years, four to five years when I started being somehow profitable. But to me, I have to say I was always struggling trading profitably discretionary because I missed some background, some data that I could, for example, use if I was in drawdowns and so on.

So when my, I started, in fact, then I started investing systematically like those mean-reversion or not, these rotational strategies like gold, ETF on SPDR and these very simple rotational strategies that are working. And I think many people could start with them just to get the confidence in systematic trading. And then like seven to eight years back, I switched completely to systematic trading.

I traded many assets, for example, mean reversion strategies on Forex, I traded breakouts on commodities, and then mainly I created broad portfolios of uncorrelated strategies on equities. And few years back I started investing in Bitcoin because to me it was interesting asset, but I don't know why. I didn't recognize the volatility and the possibility of how crypto in general is volatile, know,

because for us traders, I would say volatility is the first reason or the reason number one, we, why we want to trade the asset and that's why it, it took me like another two years before I started trading crypto systematically like two years ago.

But the problem of course is that there is no infrastructure for us algorithmic traders. So we have to start building the overall infrastructure for crypto trading with my partners at Robuxio, because the problem in crypto is that, for example, on equities, you can trade semi automatically or semi systematically which means that you will download the data at the end of the day, you will run your scanners and you will send your orders.

On crypto, it's impossible, especially if you want to trade ten, fifteen strategies together in one portfolio.

That's why we started building this backend and started trading purely automatically.

Andrew Swanscott 5:13: Yeah. Okay. Well, we're going to dig into that a little bit deeper in our discussion today.

But I just want to start with a Twitter message between you and I. This was a private one, I hope you don't mention it. I hope you don't mind that I share it, it's a good message.

Just reading it here. I'm trying to, so we're talking about, I think I might have been asking you why, what are you doing with your trading? And you said, I'm trying to give back some of my knowledge on Twitter because especially in crypto trading, people have such wrong expectations.

And it's a bit of a shame because I still think we have a nice period of a few years of many huge inefficiencies in crypto. So that's a good comment. What types of wrong expectations are you seeing people have in crypto?

Pavel Kýček 6:03: Well, there are many, many, many wrong expectations. Most of them have their roots in hindsight bias. You know, there are many coins that made like 10000%, 100000% in the past, and everyone is jumping on the last best crypto opportunity to make those 10000% on this individual or single coin, which is basically the biggest problem in crypto.

Then of course, that have their roots in this thinking, because how I feel it for us traders, once we get wrong mindset about trading then it is very hard to switch to the proper risk management type of thinking that we should always hold and we should always be in. And once you get into this gambler mindset, like buy this coin because it will be going through the moon overnight and so on, then you are basically done.

I don't think you can make any money in crypto.

Andrew Swanscott 7:13: Yeah. Yeah. That's a really good point.

And I think we're going to dig into risk management a little bit later, but at a high level, what are some of the risks with crypto trading?

Pavel Kýček 7:25: Yeah. In crypto, one of the biggest risks is just being in crypto, because yeah, it's just true, in crypto there is a lot of regulatory uncertainty still, even though there are like new laws that will be valid in this or following year, for example, in Europe, there is still a lot of uncertainty here. So one of the biggest risks is just trading on single exchange because we all know FTX case and so on.

Another huge risk is investing into one or two, three coins, not to allocate accordingly. Another risk is just being scammed by many people who are trying to sell you something or just purely steal your assets from your wallets and so on. So you really shouldn't be too concentrated if you want to invest into crypto.

But on the other side, my premise is if all these risks will be covered and they will be covered in the future, I'm pretty sure about it, potential rewards will be much lower. So we are again to this reward to risk ratios and expectations. To me in crypto, it is more about allocation, how much you should allocate rather than whether you should be allocating, because yeah, potential is still very big here.

Andrew Swanscott 9:00: Yep. Well, let's talk about that potential a little bit more. In the second part of that Twitter message that I just read out a moment ago, you said that we're going to have a nice period of a few years of many huge inefficiencies in crypto.

So let's talk a little bit about these inefficiencies or more specifically, maybe about building crypto strategies. Now I'm sure we've all heard about the huge rise and fall of some of these crypto markets. What type of strategies do you think or trading style works best for crypto?

Pavel Kýček 9:36: Yeah. This is very, very interesting question because again, my premise is that I don't know much about market, especially I don't know much about the future of the market, especially in crypto, because we all know crypto is pretty young asset. So now if I should pick up one trend, one strategy or one approach that is making the most money, it will be trend trading or trend strategies or breakout strategies.

On the other side, we don't know how the market in general will be evolving. So that's why I'm thinking about it more in general. For example, will it be behaving in the future as Forex?

Will it have more like range bound characteristics, for example? Or will it be an asset that will in fact have some value because most of the crypto coins just don't have any value at all. That's why there are very short but sharp trends, but not long term trendiness as we can, for example, see in stocks.

To me, it is really more about building broader portfolios and taking advantage of trends through trend strategies and breakout strategies and inefficiencies through mean reversion strategies. Because, yes, there are big trends that you can catch thanks to trend strats.

But on the other side, there are many of those pump and dump schemes or basically like overreactions, thanks to the amateurs that are on the market that you can trade against, thanks to mean reversion strategies.

Andrew Swanscott 11:25: Yeah. Yeah. You hear these stories of hear these stories of traders from the 1970s or 1980s.

And they say, when I was trading the futures markets back then, I could just put on a simple moving average, buy above, sell below and make a lot of money. They had these really basic trading strategies that seemed to work well. And then over time, as I guess the markets become more mature, you get more participation, different levels of experience in those markets, it tends to get noisier.

And the characteristics of the markets change, which means then you've got to adjust the style of trading a little bit. So are you seeing that kind of development now in crypto? Like do simple strategies work now or do you need to be more complicated?

Pavel Kýček 12:15: That's exactly what I expect even in the future that something like that will be happening here. But for example, let's talk more specifically. If you try some moving average crossover strategy, it will be making money on crypto.

But you can see that three, four years ago, the edge, the overperformance compared to some benchmark was much higher than today, for example. But it is still overperforming benchmark, just this very simple strategy. On the other side, mean reversion strategies, for example, to the short side on crypto weren't that profitable two, three years back, but now they are performing massively.

But of course there are big risks because crypto is exponential asset, so you can't just short these huge moves on the short side. Mean reversion strategies, I would say have to be more professional built compared to even equities, because for example, equities are like short term mean reverting market, I would say and mean reversion strategies are performing quite well.

On crypto, at least to the short side, you have to think about it a little bit more in which time and in periods you want to trade them.

Okay. But yes, it is exactly as you said, I can already see that the very simple strategies are performing well, but a little bit worse than two, three years ago.

Andrew Swanscott 13:59: So that's interesting because you can, I guess you could study the developments of these other markets that are more mature now, the futures markets and kind of overlay that on what potentially the crypto market might be doing, maybe in shorter cycles, because we've got a lot of technology and stuff these days that accelerates that type of thing? It's pretty cool to think of maybe you've got a, a, what's it called?

Like a roadmap or a path to follow in this newly developing market now.

Exactly. That's by the way exactly what we are doing

Pavel Kýček 14:31: in Robuxio and by myself when I'm building strategies. I'm looking at the past. I'm looking at commodities in 1980s.

I'm looking at, for example, stocks on the start of 2000 before the tech bubble or during the tech bubble. You know, you have to find some other assets to make your robustness testing and to see what was working in those times and in those periods, because I think it is worth it.

For example, I still don't get the reason why, for example, Larry Williams doesn't trade crypto because he used to be trading commodities, which I think in 1980s were similarly risky as crypto these days.

And this was also the reason why he was making 10,000 or he made 10,000% in one year, audited. And in crypto, I think in a very good year, can still make something like that if you want to go through such huge risks as, for example, Larry went in 1980s.

Andrew Swanscott 15:40: Yeah. Larry is not 20 years old now. Maybe his appetite for risk is a little bit different, but Probably.

Pavel Kýček 15:48: But I believe that young Larry would definitely trade crypto these days. I'm pretty sure about it. Yeah, exactly.

Andrew Swanscott 15:57: Now I just wanted to I noticed there's a comment here from Antonio in the chat, which is kind of leading to my next question about volatility. So Antonio said, check the standard deviations per hour if you scalp, which is an interesting approach. You've already kind of mentioned the volatility in crypto markets can be pretty huge at times.

Use How do that knowledge or that insight to adjust how your trading strategies work? Yeah.

Pavel Kýček 16:26: Well, first of all, I love the volatility of crypto. I think every trader basically loves volatility. To me, it's the biggest advantage of crypto to tell you through.

It's the reason why I'm in crypto. So I'm dealing with the volatility is that I'm trading portfolio of uncorrelated strategies. And it means that I'm also trading strategies in which every strategy basically trades up to fifteen, twenty positions.

So this is first thing how I'm dealing with volatility. Many uncorrelated strategies, every strategy has a little bit different role in the overall portfolio. And that's why I'm also having some maximum threshold how much exposure I want to have to one coin, for example, because as I said earlier, one of the biggest risks is to be overexposed to one coin because you never know if this coin will go to zero overnight.

So that's why the exposure, the overall exposure to one asset, one coin has to be pretty low. And then I'm really dealing with the volatility the way that I'm trading mean reversion strats, trend strats, breakout strategies, and I'm spreading the capital accordingly.

Andrew Swanscott 17:50: Yep. Yep. Now you make a good point there about some coins potentially going to zero and some have had a huge run up.

And you made a point in, I think it was two weeks ago in the trading panel about survivorship bias. Or maybe I made the point. I don't recall.

We were talking about it anyway, right? Yeah, we were talking about it. Yeah.

And I noticed when I was looking on your blog that you had a really interesting chart about survivorship bias. And do you want to talk about this one in crypto? Think this was a really excellent example.

I'm going to put it on the screen. Yeah. Yeah.

Can you talk about the impact of survivorship bias on crypto trading? Here we go. It's coming up.

Boom.

Pavel Kýček 18:33: Yeah, well, survivorship bias has huge impact to any trading, not just on crypto but especially on crypto, the impact is huge. Why? Because crypto is firstly exponential assets.

And secondly, there are many losers and just a very few of real winners. So for example, this example is very good one. If you trade, for example, today's top 10 coins with some trend strategy, which is basically this example, you would get the performance of the blue chart.

But if you would be trading top 10 coins on the day when they were really in top 10, you would get the green chart, which is basically for like 5x lower or maybe even 10x lower the performance. So the more volatile the asset and the more immature the asset, the more the survivorship bias can affect your trading and your backtesting, especially your backtesting.

Because if you don't backtest properly on survivorship bias free database, then you just can't get the proper results that should somehow be similar to real life trading.

But again, this is quite problem because for example, for equities, you have your Norgate data, but in crypto, there are not many services. I don't know about any. So that's why we built our own solution just to have proper data and proper results of the backtests.

Andrew Swanscott 20:19: Yeah. Yeah. I think this is a really excellent example that shows the impact of survivorship bias.

I remember when I was trading stocks many, many years ago and I had access to the Norgate survivorship bias and the results can sometimes be huge. This is an excellent example because I think a lot of people, especially beginner traders, don't understand the impact that can have. And so they would think this blue line is what they're going to trade and then they get a bit of a shock when it's nowhere near what they're expecting.

Pavel Kýček 20:49: This is a brilliant example here. Yeah. Thank you.

But the survivorship bias is in fact everywhere. You can see it even on ETF investing, that people are choosing just for example, NASDAQ because it is over performing these days or they are choosing, I don't know, Nvidia because it is performing like crazy these days, but you never know if it will be over performing even in the future. So that's why this survivorship bias and hindsight bias

Andrew Swanscott 21:22: are two biggest problems for traders and investors in general, I would say. Yep. Okay.

Now I want to move on to robustness testing in a moment, but I know this question is, it hasn't come up in the chat yet. I'm quite surprised, but people will ask, what platform do you use for testing? Where do you get your data?

All that type of practical things. So how do you actually do your testing, your strategy Well,

Pavel Kýček 21:48: first of all, you have to have good data, which again, in crypto, it is a bit of problem. We are getting data from Binance and from KuCoin. We are creating and building our own survivorship bias free databases on which we are making tests.

For testing, we are using RealTest, which is a software I can really recommend to anyone, especially to non programmers as myself, for example, because the code, the language is like super simple. I can compare it for example to Amibroker that I used to be using in the past and RealTest is really very, very nice and very simple, very logic software.

And then of course, I think that if you really want to be serious with your trading, you have to have at least two solutions that you cross check your results between each other.

So that's why we also built our like proprietary backend backtesting software. And we are like checking those results with RealTest and we are having like two different source of backtesting. But real test is good start.

The only problem is the data. I don't think you can find proper data for survivorship bias free testing these days, or I didn't find any to tell you through.

Andrew Swanscott 23:19: What about the length of the data? Because some of those coins don't have a lot of history, right? Compared to if you look at stocks or the futures markets, go back How a long does that impact what you're testing?

Pavel Kýček 23:30: Of course, history, that's true. We just don't have enough data in crypto. That's definitely true.

That's why it's affecting the roots of building the strategies. I would say you really have to think about building strategies a little bit differently compared to, for example, commodities compared to equities where there is a lot of data. So that's why, for example, I'm building the strategies that way that most of my strategies have some roots in very old strategies.

For example, from Larry Williams 1980s or another well known traders from the start of 2000, because I can see that those strategies went over quite long out of sample, like real testing. Of course you can't use the strategy by itself for crypto because crypto is much quicker, but the logic, the entry logic or the exit logic can be very similar to those like older ones, but you have to spread the capital.

You have to really trade every single strategy as a portfolio And it is like sub portfolio for your overall broad portfolio because otherwise you are just too exposed to one coin.

Because basically the idea of my trading is making money from volatility, not from betting on one or two or even 10.

Andrew Swanscott 25:03: Yep. Okay. We've got a question in the chat about diversification, which we'll get to a little bit later, because I know you've got some good insights about that one.

But let's jump to robustness testing for a bit. I know this is one of your favorite things based on our previous discussions and also on your blog. You talk a lot about robustness testing.

So how do you know when a

Pavel Kýček 25:28: strategy is robust? What are you looking for? Yeah, well that's a very good question to tell you through.

I have to say that you never know if your strategy is robust, know. Like really, you really never know. You just have to do as much as possible and go through some robustness procedure and through robustness processes to know that you did as much as possible for the strategy before letting it trade live.

But before even robustness, you have to make really one step back and you have to build the strategies properly. For example, my ideas are that you have to build the strategies with idea first approach, which means that you follow some logic. So what you want to trade, if you want to trade trend strategies, mean reversion strategies and so on, because then you can go to the proper robustness testing.

For example, it is quite hard to use some walk forward optimization and walk forward robustness testing because on five, six years of data, you can get very good results just by pure luck. So this is not good method in my opinion. What I'm looking at the most is really robustness over parameters,

which means basically that, for example, if I'm trading some very simple strategy and I'm using moving average 10, for example, moving average 20, I want to see that the strategy is performing very similarly with moving average, five, seven, eight, up to 15, let's say minus 50% plus 50%.

Another thing that, or another approach that I'm using is that I'm testing all our strategies on four hour charts or twelve hour charts just to get as much data as possible. And I want to see them performing well or at least average, I don't want to see them like going down immediately.

And another thing that we are using just to get more data for example, I don't think that many traders are using it is that we are taking the advantage of crypto being 20 fourseven market, which means that we are creating our own artificial daily closes.

So we are using the official daily close and we are having another 23 unofficial artificial daily closes. And I want to see again the stability over these artificial daily closes and how it is impacting the strategy in general. So this is one of robustness testing,

but another one I would say that is having its root just in pure logic is for example that if I'm trading trend strategy to the long side, I want to see this strategy performing well in trends, basically in long trends.

I know this is very simple logic, but not many is using it. If I'm looking at mean reversion strategies through short side, I don't mind if it is losing money in strong long trends. And if you really start building your strategies with idea first approach, you know when your strategy should be making money and when it should be losing money.

And if it is losing money in the phases when it is supposed to be losing money, then you don't have to care about it that much. You just look at it and you are basically looking at like general parameters and if it is still in some range.

Andrew Swanscott 29:20: Right. Okay. So what about which market do you actually test on?

Like, obviously the big one is Bitcoin. Are you testing strategies on Bitcoin and then testing them on other markets as well? Or do you like, where do you start your testing?

Pavel Kýček 29:35: I never test the strategy on single coin. I'm always testing my strategies on survivorship bias free databases. And I'm always starting with, I would say some kind of basic characteristics I want to trade.

For example, I'm building the strategies that way that they have to be able to trade with few mills, for example. So I need to be able to trade on highly liquid or at least average liquid coins. That's why I'm building the strategies that way that I will choose some data set of highly liquid coins and I'm testing the strategies on those coins, but without survivorship bias free.

So let's say I want to trade on 10 most liquid coins, which I don't want to, I'm choosing a little bit different parameters but let's say 10 most liquid coins and you have to use the same 10 liquids coins today or the 10 most liquid coins like one year back but not the same ones. Just have to have this survivorship bias free database, which is really again the biggest mistake I can see in beginner traders

but even in intermediate traders, would say not many are realizing how big the survivorship bias free, like survivorship bias is.

Andrew Swanscott 31:10: Yeah. So you made a comment in the previous answer about looking at the performance results of a strategy when it should lose. Yeah.

So do you ever test with regime filtering or trying to switch a strategy off? Because I guess using this idea first approach, which you talked about, you kind of know logically when a strategy should work and when it shouldn't. So does that lean nicely into regime filtering?

Pavel Kýček 31:38: I'm using regime filters with some strategies, but not with all of them. It again depends on the nature of the strategy. For example, to me, it doesn't make sense to use trend strategies without regime filters.

Why should I enter into long trend positions if the overall market is falling like crazy? And the same to the short side. So that's why I'm, for example, using regime filters for trend strategies.

I'm not using regime filters for mean reversion strats that much. It depends again. But again, what I want from my trading solution is to be as diversified as possible.

So that's why sometimes I don't use regime filters, even though with regime filters the performance would be better, but I will get slightly different logic from the strategy. So that's why sometimes regime filters are good, sometimes they are not.

Andrew Swanscott 32:47: Yeah. I guess though, as we spoke about a little bit earlier about the limitations of the length of the historical data, you also get, I guess, a limited number of different types of regimes. Does that reduce the reliability of regime filtering?

Pavel Kýček 33:06: This is very good, very good question. I think it depends on what type of regime filters you are using. Because again, if you are using very simple regime filters, those that for example, you can use on equities, something like, let's say very, very well known regime filter is 200 moving average on S&P or something like that.

And you can reuse this kind of regime filter on crypto, for example, on Bitcoin or on TOTAL3 crypto index, it doesn't matter that much but I'm pretty sure that this logic is very strong and you can use it even though there is not enough data in crypto. But I'm really not trying to be like super crazy with regime filters because there could be this issue that you mentioned that you can just get good results by pure luck.

And this is not something I would like to get.

Andrew Swanscott 34:10: Yep. Okay. Let's move on now to portfolio and diversification a little bit.

It's come up a few times in the chat here. So first of all, you've of already touched on this a little bit already, but maybe might be good to just review it. Why do you think it's important to have a portfolio of strategies in crypto or I guess in any market really?

Pavel Kýček 34:33: I think it depends on what you want to get from your trading. For example, if you want to maximize profits and you don't care about stability of your returns at all, for example, you can bet everything on one single trend strategy in crypto and maybe or even very likely you will make the most money in this, let's say bull market if this bull market will continue.

But if you want to get stability and you want to be prepared for almost anything in markets, you just have to spread the capital between many different approaches just for you to know that if the market will be evolving — and it will be somehow evolving, we just don't know how — you just have to have many different trading strategies, because then you are basically covered.

My general approach is trading these all weather portfolios even on equities, on crypto, because that way you can make some money if the market will be, for example, highly volatile but it will be moving basically sideways with mean reversion strategies or you can make money if the market will be moving up or down, thanks to your trend strategies.

And then the only phase that is really not good for us directional traders is basically sideways low volatility market. This is the market where you will lose money probably.

Yeah. This is what it is.

Andrew Swanscott 36:19: Yeah. We've got a question in the chat from Dalibor. Apologies if I said your name incorrectly.

This one's about diversification, but I like it because it's in the context of correlation. How do you diversify in crypto with everything so correlated?

Pavel Kýček 36:34: Yeah, this is good question. You know, it is also about the meaning of diversification. Of course, if I want to think about diversification in general, then I think that you should be allocated into crypto but probably you should diversify into different assets.

So this is I would say the highest level of diversification. If we get to crypto you can of course diversify because you can just go over top 20 coins these days and go one by one and you can see that some of them made like today, let's say 5%, another -12%, another one -6%. So yes, they are diversified long term, but if you are trading, because it also depends on what type of strategies you are trading.

I'm trading shorter term strategies like one day, two days, but also weekly strategies where you are holding your position two, three weeks. And especially with these quicker struts, the diversification plays a big role because you never know when you will hit those outliers that will really push your equity much further. So yeah, diversification.

I get this question pretty often, but I don't agree to tell you through because it is more about the general approach of trading. Because many times people are thinking about diversification and trading the way that you are trading. Trend strategy on Bitcoin, mean reversion strategy on Ethereum for example, another trend strategy on, I don't know, Dogecoin or something like that.

And then you are right that the correlation would be pretty high, but if you are trading like a portfolio in every strategy, then the diversification is pretty big.

Andrew Swanscott 38:30: Yep. Now on the trading panel, I think it was two weeks ago, you made a comment, which we didn't really dig into a lot at the time, but I thought it was interesting and I saved it for today. So you said something about you really went into in your crypto trading, you really went into oversimplification of strategies and more, I think you said crazy but advanced techniques in the portfolio.

Can you expand on that a little bit more? What you what exactly do you mean there? Yeah.

What I mean by that? Well,

Pavel Kýček 39:03: firstly, why I'm trying to oversimplify single trading strategy? The main reason is that I don't know if the strategy will be performing in the future and the more metrics the strategy will have or more conditions, the higher the probability that the strategy will fail in the future, that it was just overfitted strategy on few years in the past.

So that's why I'm really trying to build a strategy that has one, two entry conditions, one regime filter, and let's say one exit condition, something like that.

Of my strategies have just one entry condition and one exit condition, right? That simple. But their main logic is very strong.

And this is how I'm trying to build the strategies because as you said, Andrew, we just don't have enough data in crypto. So that's why the strategies really have to be super simple just to be at least somehow sure that they will be performing in the future. And by the advanced approaches in portfolio building, what I was thinking about was that you really have to think how you are building the portfolio,

because I can see traders that are like building many strategies and they are just putting them into the portfolio just for the purpose to have as many strategies in the portfolio as possible.

To me, it's pure nonsense because that way, for example, you can be good for, I don't know, trend creating trend strategies, for example, and you will build 10 trend strategies and you will put them into the portfolio but how much they are diversifying your overall trading results? Not much, not much.

That's why I'm trying to build or add every strategy to the portfolio only if it will have some diversive, know, only if it is like diversifier to the overall portfolio.

If not, I will just keep it somewhere, but I won't use it for the portfolio building in general. So this is really something that's why I can imagine that you built your very simple, I don't know, crypto strategy in one day, but the portfolio, the proper portfolio, you will be building many months, maybe even a year. And in fact, you are still working, maybe even you are still working on your portfolio to be as good as possible.

So this is how I'm thinking about it. Yeah.

Andrew Swanscott 41:56: So how do you then determine if a strategy is contributing to the diversification of the portfolio? What are you looking at? Yeah.

So firstly,

Pavel Kýček 42:05: basic logics, which means if the logic is the same or very similar to any other strategy that I already have in the portfolio, and then I'm using as probably everyone correlations, correlation coefficients. I'm looking when especially I'm looking at drawdowns, which means when the strategy is creating drawdown compared to similar strategy that I already have in the portfolio.

And if the strategy is building the drawdown in different period just by luck, or if it is by some characteristics that is in the strategy, if you know what I mean.

Really for me, is not about trying to push the performance as high as possible, but trying to push the stability as much as possible. And the only thing that you can basically control is drawdown, not performance because performance is more function of the trendiness and of the volatility of the market. But the drawdown is function of your portfolio or strategy building.

Andrew Swanscott 43:18: Yeah, well said. I think one of the appeals, especially for beginner traders is when they're building a strategy to make it look like a really nice straight line. I imagine in crypto, you can probably build ones that go exponential, right?

Depending how much you could have fit it. But by the sound of it, you almost, you probably don't even look at the equity curves, right?

Pavel Kýček 43:42: Well, I'm looking at equity curves, but I'm not trying to build the best equity curve possible because again, I'm thinking about every single strategy as part of the portfolio. I could never imagine trading just one strategy. And I also think that it is big, big problem for many traders that they are trading one, two strategies because they just fall in love with them then and they try to over optimize them.

They try to overfit them. They really try to make as much money with just one or two strategies instead of building more uncorrelated strategies in broader portfolio. Think this is the biggest pain point of many traders that they are really trying to build the best possible one single strategy which can be working.

I can imagine that it can be working all the time because you have trend strategies, have mean reversion strategies and breakout strategies and every strategy, every trading approach makes money just in some market phase, not in every market phase. So this is how traders should be thinking about it.

And I think you had a very good point about building strategies on the past because this is another huge trap like novice or beginner traders are falling into that they are trying to be rich on the past data.

Don't think about the market that way that future will be always very different compared to the past. It will never be the same and you should have it always on your mind if you are building the strategy that past is really only the past and the future will be probably very different and probably worse.

Andrew Swanscott 45:33: Yep. Okay. Well, we've got some questions in the chat, which I want to get to in a minute, but just as an extension of what you were just saying, what other, I guess mistakes do you see traders making when they're approaching Algo crypto trading?

Pavel Kýček 45:50: Well definitely overfitting. It's number it's number one problem of all then they are building the strategy of one asset, especially in crypto. You can see it all the time.

Everyone is selling those bots that made 100% in last three years just because it is trading or it was trading Solana that made like 100,000%. If you have such an exponential coin, then everything will be making money on it. This is really a huge trap that I can see all the time how people are falling to it.

Another one of course is building strategies on like survivorship bias or on data that includes survivorship bias which is very big especially in crypto or in equities. For example on commodities is it is a little bit different. This is not such a big deal but on equities and crypto it is a huge problem.

Hindsight bias is a huge problem. And then in general, I can also see that algo traders are making just mistakes in quotes, that they are not checking their codes well. Yeah.

Even I made quite a few of those mistakes in the past and I even trade it unfortunately on live data. I'll I'll I really did. So, everyone did it probably but this is big problem too, that we think that we got the code and we just let it right and yeah, and we just lose.

Andrew Swanscott 47:30: Yep. Okay. Well, we've got a couple of questions here in the chat.

So let's go first here with John. So this is more of a clarification one. And this actually came up when you're talking about regimes.

I just didn't see it on the screen at the time. So John would like to confirm. So in a long regime, you'd limit momentum breakout trades to long only, but mean reversion and trending strategies are free to do their thing.

Pavel Kýček 47:56: Yeah. Well, it depends. It depends.

Some of our mean reversion strats do have regime filters because for example, the logic is in for regime filters. For example, let's think about some correction. Want to trade correction to a trend.

Trading correction in general is mean reversion trading. Why shouldn't I use regime filter when I want to trade correction into trend? In this mean reversion types of strategies, I'm using regime filter, which is basically one of the conditions of the strategy.

But it really depends. It is not that simple that I would say mean reversion without regime filter and trend strategies with, because I try to combine as many different approaches as possible.

Andrew Swanscott 49:06: Okay. Thank you for that. Question from Antonio.

Do you focus on mean reversion from a range expansion perspective?

Pavel Kýček 49:18: Mean reversion from a range expansion. Yeah, well, if I understand the question correctly, then some of our strategies are using range expansion condition for mean reversion strats but again it depends. I think I think this is very good question in general because what I think is one of those problems you were asking Andrew what traders are making like mistakes when building struts is those indicators versus price action wars,

because to me you can describe a range expansion with indicator or with price action pattern.

It doesn't matter in fact, this is just the tool and how you describe it is up to you, but how you trade it, what you want to trade, this is like one step back that one should be always thinking about before building the strategy. But in general, range expansion, playing with range expansions for mean reversion strats, breakout strats is good way to go, not just in crypto, but in any highly volatile asset.

Andrew Swanscott 50:38: Okay. Here's a question from, my goodness. I don't know how to say this.

Looks Croatian maybe. Can you say that one? Sorry for butchering your name.

Does your trading system analyse order books and do more complex decisions on positions? Or do you focus more on logic, the strategy logic and care less about exchange mechanisms like market orders?

Pavel Kýček 51:02: Yeah. I'm more concentrating on really general logics on higher, higher data like daily, weekly and so on, because you have to take, you really have to think about cost of trades which is another big mistake traders are doing that they are not thinking about cost of trading which is not only fees but also slippage and so on.

And yeah, the lower you want to go and the, for example, here order books and this is really quick trading and I don't feel comfortable trading this kind of strategies on crypto because again I think that especially those very quick strategies will be changing the characteristics very quickly as smart money will be entering crypto markets and so on.

So that's why I prefer like trading slower strategies because even on stocks or on commodities, you can see that those strategies, if they are losing edge, the face of the edge losing can take even a year or so. But if you are trading one minute, two, three minute strategies, I have seen those strategies that they were performing and in a day, two or a week they were just going through zero very quickly.

So that's why especially in our surveys I want to have the stability pretty high, so that's why I'm really concentrating on this daily, weekly or maybe 12 strategies.

Andrew Swanscott 52:44: Okay, thanks Pavel. Another question here from John. Are you trading CFDs or actual crypto or something else and why?

Pavel Kýček 52:54: We are trading futures on crypto and why? There is biggest liquidity and lowest fees and basically this is the best market for us traders. Yeah, and it is also the simplest solution for trading.

So there is not many reasons to trade differently if you can trade that way. For example, I know that in The US you can't trade futures on crypto. So this is a very different question then.

But if you can, I think futures are the way to go?

Andrew Swanscott 53:35: Yep. Okay. Now there was a question submitted on X to you from Sean.

It says, I can't put this on the screen. I'll just read it. If you're a US trader who cannot trade crypto futures, would you trade actual coins including smaller ones or just FBTC LTCN?

Pavel Kýček 53:57: Yeah.

Andrew Swanscott 53:58: The biggest Keep slippage and commission low.

Pavel Kýček 54:01: Yeah. This is good question but very hard to answer to tell you through because there is a problem with the universe in general. I'm not a big fan of choosing five, six, seven, eight or 10 coins and trade just them because again, you are getting into some kind of a hindsight biases.

But if I would have to choose, I would trade probably the top 10 biggest coins with trend strategies, because trend strategies have the highest average trade and that's why even fees and slippages on like general crypto market is not such a big problem.

Andrew Swanscott 54:49: Okay. Question here from Summer Squad. Welcome Summer Squad.

What's the simplest system or method you would advise a beginner to start with that would work across commodities too?

Pavel Kýček 55:04: Yeah, that's okay. Good question. If I should choose the simplest system, I would choose something like Donchian — I still don't know how to pronounce it — some kind of Donchian breakout strategy like thirty days, one hundred days or something like that and really stick to breakout/trend strategies type of approach.

Probably the most robust one.

Andrew Swanscott 55:39: And then a similar kind of question here from Mr. C. Any reading recommendations for finding trading system ideas?

Pavel Kýček 55:47: Well, of course my Twitter account, but yeah, no, I'm trying to share quite a lot of information. But if I would start building crypto trading strategies from the scratch now, I would really go over old books, like really Larry Williams books, Linda Raschke books, those type of traders that were trading and writing books in times when commodities were really moving like crazy and you can definitely take some kind of inspiration there,

but you also have to add some books from, for example, Nick Radge and so on, just to get the feeling of what portfolio trading is because you just can't trade crypto without portfolio approach. I think this is too risky.

Andrew Swanscott 56:44: Yeah. Yeah. I think another good book it's probably worth looking at is Perry Kaufman's.

I think it's called Trading Systems and Methods or Yeah. Something like Which was a was an older book, I think you might have updated it. That's got amazing trading ideas in it.

So I imagine a lot of those would work in crypto too.

Pavel Kýček 57:03: Definitely.

Andrew Swanscott 57:05: Okay, here's another question from Mr. C. If you use Monte Carlo simulations with random data, parameter sensitivity testing and all the other robustness tests you mentioned, is the system ready to go live for live trading?

Pavel Kýček 57:21: I don't know if I understand the question properly, if this is the only robustness testing for crypto, but for me, especially with the amount of data we have in crypto, this is not enough because really the robustness testing starts with how you build the strategies. You just can build your strategies on data.

I would say indicator first type of strategy building and just go over the general robustness testing that are very usual in traditional finances.

I would say this is not enough because if you make Monte Carlo on five years, six years and two huge long trends, what kind of information you get? The information is really, the value of the information is pretty low. So that's why this is definitely not enough.

You also, I think in crypto, especially in crypto, you already have to know how to build the strategies properly because yes, I'm repeating all the time that I'm building very simple strategies but sometimes it is quite hard to build simple strategy that will be performing in the future too because you really have to understand the markets. This is the main idea I think.

You really have to understand the market logics, the market dynamics and if you do, then you can build a profitable strategy on any type of market.

Andrew Swanscott 59:00: Okay, we've got a final question here from John. Have you ever looked at a system of closing a portfolio's complete set of trades at say 1% gain? This short circuits mean reversion multiple position trade sets.

Pavel Kýček 59:23: Not sure if Yeah. It I unders

Andrew Swanscott 59:25: kind of goes against the trend following principles though, doesn't it? Yeah. Closing out for trades early.

Closing out for early profits.

Pavel Kýček 59:33: Like, in general, I'm not using profit targets at all. This could be an interesting question too, because I made so many tests on equities, on commodities, on crypto and I almost always find, if not better performing, then definitely more robust approach than using profit targets. And the same is valid for stop loss.

In fact, too, you can manage the risk differently than with stop losses. So yeah, I'm not doing something like closing positions if my portfolio is plus 1% or so. I think you just don't, you can't use the potential of crypto or of the asset that is moving in fact, if you set some boundaries with let's say 1%, 2% or so.

I think this is discretionary type of thinking in general that is not profitable even with discretional trading, but definitely not on algo trading or I haven't found this approach at least. Yeah. John clarifies

Andrew Swanscott 1:00:44: reduces mean reversion drawdown in that case. I think you answered the question. Sorry, do you have more?

Pavel Kýček 1:00:53: No, no. I just don't get the question how I should, but yeah, I don't know if you understand it, Andrew. What's

Andrew Swanscott 1:01:04: this question? No, think you covered it pretty well. Yeah.

Thanks for the question, Don. Thank you. All right.

We're just about to wrap up here in a moment. But first it's crystal ball time, Pavel. I'm not going to ask you about the price of Bitcoin next year.

There's plenty of people who have guessed that and got it wrong. There's a question here from Sama first, then I'm going to ask you one about the future of crypto. So Sama would ask, do you think we're entering into a bull cycle on crypto assets with Ethereum and Bitcoin moving near all time highs?

Pavel Kýček 1:01:39: Good question. Unfortunately, I don't know. I wish, but really, I don't try to analyze the future of crypto market that much.

So really don't know.

Andrew Swanscott 1:01:53: Yep. Okay. And then so my question is not really price based, but what do you think is the future of crypto?

We're seeing talk about different types of regulation. Some countries are making it illegal. What do you think is going to happen in the space?

Do you even care?

Pavel Kýček 1:02:11: Well, I do care. I do care because I used to be working in traditional finance in the past too. So I'm having quite a few contexts and I can see how it is evolving.

I can see how even banks are interested in crypto. Why? Because they can see their profits of course, because the margins, the fees are much higher with crypto than in traditional assets.

So what I think that will be happening in the future is that there will be more and more regulatory clarity in the crypto space and that's why more and more smart money and bigger funds will be entering this space. And that's why these inefficiencies and volatility will be getting lower and lower. Now I'm talking about like five years or so in my opinion.

So what I would expect is that over the time, like in ten years, five to ten years, crypto will be just another asset that will be like diversify for your portfolio, but not this huge opportunity that is right now. Yeah.

Andrew Swanscott 1:03:21: All right, Pavel. Well, thank you very much for our chat today. It was really informative and we've got a lot of great questions in chat as well.

So thank you very much for that. Now, how can people learn more from you or maybe even get in contact with you?

Pavel Kýček 1:03:37: Yeah, thank you. Thank you, Andrew, for having me here. So let's just, you can go to robuxio.com or to my personal Twitter Pkycek, P K Y C E K.

I can see you mentioned it here on screen. Or just write me — I'm on Twitter quite often, so I'm always happy to answer any question.

Andrew Swanscott 1:04:04: Yep. Excellent. And I'll put links to those in the description of the replay as well.

So if people want to find those, because that's a difficult one to spell in Twitter. Thanks again for your time today. We've got some comments in the chat, which I might just share here quickly.

And then, so C says, thanks for the podcast. Summer said, thanks for the great info. I'll have to rewatch as I missed out the beginning.

And actually John put a comment at the beginning, which I haven't put up on the screen, which I'm going to do now because it was a good one. John said, I really appreciate Pavel's approach of mixing trend, momentum breakout and reversion to mean strategies. Thank you, John.

Explained a lot of that today. So yeah, thanks again for your time today, Pavel. Any closing comments or thoughts before we finish up?

Pavel Kýček 1:04:57: Well, think about trading crypto because I think there are many risks, but there are many opportunities, huge rewards. And at the end of the day, is about capital allocation, not if you should be trading, but how much.

Andrew Swanscott 1:05:14: Yep. Very well said and a nice way to end the show. So thanks again.

Thanks to everyone for joining us and we'll be back again. Oh, come to the trading panel. It's every week on what time is it?

I always forget. 4PM. Pm Eastern every Friday on all the Better System Trader channels and Pavel comes along as well. So you'll see him there again, I'm sure in the future.

So thanks again, everyone for joining us and happy trading. Happy trading.