Generating insane returns with quant crypto trading

Transcript

Pavel Kýček 0:01: In 12 models which are trying to catch the edge and models which are hedging your portfolio. Let me give you an example. For example and then we have models which are regime really regime dependent, and these are only, let's say, switched on anytime the market starts doing its thing.

You know? It's starting really moving to the long side a lot or to the short side a lot. You are able to get the best of both worlds, which means a higher win rate, which you are getting through mean reversion and also the exposure to the momentum effect, which you are getting through momentum strategies.

Show intro 0:47: 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.

Tristan Pollock 1:20: Alright, Pavel. Welcome to the show. It's so great to finally be chatting crypto on the Algo Advantage.

Welcome, mate. Good to see you.

Pavel Kýček 1:29: Thank you. Thank you, Tristan. It's nice to be here.

Tristan Pollock 1:33: Awesome. Let's dive straight into it, Pavel, because we'll probably end up talking for hours and hours otherwise. So really good to finally talk crypto on the show.

we've maybe touched on it a little bit here and there, But let's face it, it's a huge realm for quant traders at the moment. There's lots of opportunities there. The market has matured enormously.

Things have changed globally. I mean, I think US citizens now have even got a lot more access than they used to. So, it's definitely overdue for us to chat about the opportunities here and how we can take what we know about quant trading if we're not already trading crypto and apply it to the crypto markets.

And I guess the first thing that I thought of is that, the whole crypto industry has brought a lot of new traders to the market, right? There's a lot of people who have gotten interested in crypto and now they're interested in the markets and maybe they've even moved from crypto into TradFi or trading both. But there's a lot of people who discovered crypto and they've taken more of a buy and hold approach.

Lucky you if you bought and held a long, long time ago, but what happened is people started to trade and even the long term buy and holders, they saw certain huge peaks and thought, Hey, I'm gonna take some profit. Or they bought a bit later and then had some huge draw downs and decided that they would sell and buy later. So it started to make traders out of investors, I think.

And so I was just curious to start off with, like if you were one of these kind of longer term buy and hold crypto guys who wasn't really super active and didn't really want to be too active in the market, but wanted to do a bit better than buy and hold. What's the first step toward becoming a trader maybe that they could make?

Are there some basic strategy principles that they could apply to, just to do better than buy and hold, but they've got their Bitcoin and they wanna make a few smarter decisions than they're doing.

Pavel Kýček 4:10: Yeah, good question. Not as simple as it can seem on the start because firstly, the disadvantage of crypto is that most of the assets, most of digital assets are not building long term trends, which we can see in TradFi. And by far, most of the new crypto coins are not reaching like, new all time highs between two different cycles.

So I would give probably one advice. If you want to buy and hold, go with Bitcoin, maybe Ethereum. Honestly, I don't know.

But crypto asset is really crypto asset class is really for trading. It's much better for trading than for buying and holding. And simple simple advice would be focus on momentum or trend following to the long side.

It would be like simple simple advice, which would have to be a little more specific because even with this very simple advice, they wouldn't have to get that far because we can see behavior which is similar to stocks that the longer the bigger projects tend to build longer term trends. The smaller projects tend to really make their pumps and dumps, and that's more or less it.

So you can take advantage of it, but you already have to have a little bit more advanced solution.

For simple solution, I would say focus on trend following on top 10 to top 20 biggest coins by market cap, And that could be a good start.

Tristan Pollock 6:06: Okay. So what else do we need to know specifically about crypto? If we're going to start trading crypto, there is obviously some significant differences and some significant risks to trading stocks and futures.

And some of the things that come to mind too, you haven't really done this before, is there's a few other key features, right? You need a wallet or you need to transfer money or you need to go from fiat (so-called) into crypto. There can be different ways to transfer money and different costs associated with that.

So like what's Do you need to get started with some real basics of just understanding what the ecosystem is and how it works?

Pavel Kýček 6:56: Well, are basics which you for sure have to cover. On the other hand, crypto exchanges, they are really trying to make their job in a sense to onboard their clients as soon as possible. So with most exchanges, it's really working that way that you send payout to crypto exchange, you change to stablecoin and you are ready to trade.

And then it depends if you want to trade to the long side or if you want to short too. And then based on it, you need to use different kind of products. You can trade on top of Crypto Spot, then you are really buying and holding the coins.

But what's better solution is trading on crypto futures. Because with futures, with perps, you are able to go long, you are able to go short and fees are much smaller compared to buying on spot market.

Tristan Pollock 7:56: The liquidity in the futures is that more or less equivalent to the spot?

Pavel Kýček 8:02: No, no, the liquidity on futures is by far higher, is anywhere between three to 10x higher because most of the institutional investors, traders, clients are trading on futures. So there are many more contracts built on futures compared to what's holding on the spot market.

Tristan Pollock 8:27: Right. And now as a systematic trader, you obviously need to get as much high quality data as possible as well. What are the challenges around getting historic crypto data and how do you get it?

What are some of the considerations for getting good data?

Pavel Kýček 8:47: Yeah, data is for sure crucial. I don't think there is a proper service right now for providing good data or at least I haven't dig deeper in last few months, let's say. So you can definitely download the data from Binance through API, for example.

That's good start, but you have to clean the data and you always have to be sure that you are building survivorship bias free database because biases in general in trading are super important topic but especially in something like crypto which is super dynamic asset where many assets or coins are really burning and being died in a few weeks or months, survivorship bias is huge here.

So you always have to focus a lot on proper survivorship bias database. And of course, you have to clean your data a bit.

So the best approach is to download more data from, for example, OKX, Bybit or other data streams, compare them, clean them. It's a process. It's not that simple.

And if you really want to rely on properly managed data, it takes some time and procedure to have this database.

Tristan Pollock 10:13: If you just download data from an exchange as at today though, you would only get the coins that are currently listed. They wouldn't provide you delisted coin data. Is that right?

It depends. It depends on Binance. You can get even the listed universe.

So you can get basically the whole tradable universe

Pavel Kýček 10:31: anytime, not just based on today's data, which is nice, but it's still not cleaned. You can still have to make some readjustments and some changes in the data.

Tristan Pollock 10:45: Yeah, right. There's obviously not a lot of crypto history, so there's not a lot of data available, but we'll get into that. What else is different about crypto?

What else is the same about crypto? So compared to trading stocks, futures, are there some other key differences, things that just really you need to be aware of? It does not work the same way.

Pavel Kýček 11:15: I would start with what's the same because then it's really much simpler for someone who is trading somewhere else to imagine what's the procedure behind building profitable trading models on crypto. I'm always trying to compare crypto to different asset classes in different stages. For example, if you want to look at stocks, crypto is similar in a way that it's more biased to the long side, and it's a mean reversion long asset class basically.

Dips are being bought most of the time oversimplified. On the other hand, in terms of volatility, crypto is more similar to, for example, commodities because they are really having these explosive, explosive movements which can come out of nothing very quickly or to smaller, smaller stocks. And another good comparison is that if you want to take some period from history when crypto was or stocks were similar to crypto.

You can look at, for example, tech bubble before 2000, when really tech stocks were having similar characteristics to what we can see in crypto. I think it's pretty logical because we are always moving in some kind of immature to mature asset cycle on any asset class in general.

And it's all about being less liquid, more volatile, having bigger inefficiencies, bigger edges in the market, which is basically bringing in bigger institutional clients, which want to extract these edges and make money out of these edges.

And it's adding the liquidity up and it's lowering the volatility. So basically, if you want to make your research on crypto and you want to take some out of sample data, not from crypto, I'm always encouraging people to go into bigger history on commodities and on stocks. Because in those times you would be able to get different edges, which were right now in those markets and similar edges to which are in crypto these days.

Tristan Pollock 13:41: Obviously, the crypto markets are maturing a lot. And in the early days, there was a lot of money to be made by far more risk free or risk neutral strategies of arbitraging across exchanges and that kind of thing. I believe you are primarily a directional trader, correct?

You're not really involved in those arbitrage trades?

Pavel Kýček 14:08: Yeah, we are building some solution for base layer where we would be able to make a few percentage a year based on arbitrages, but mainly our exposure is through directional strategies because this is really how you can extract the biggest profits in crypto.

Tristan Pollock 14:29: So a lot of those inefficiencies, would you say they have dried up now?

Pavel Kýček 14:36: Well, what we can see, especially in the field of institutional crypto traders is that the edges across arbitrages are getting smaller. And it's it's very natural because the more liquidity is getting into these strategies, the smaller the edges are becoming. But honestly, that's the same even with directional strategies.

The edges will be smaller. You would have to be prepared for it. And let me give an example.

If you would run trend following strategy anytime between 2017 and 2020, you would make on average anywhere between 100% to 300% to the long side only on a yearly basis. This is something which is not doable these days. We are still in higher dozens of percent territory per tradable model, but the edges are getting small.

That's just the fact.

Tristan Pollock 15:34: I guess the other big thing I was waiting for you to say in terms of the differences with crypto are the risks. And I know that I got burnt once or twice in crypto. The main coin or the main time that I got burnt in crypto, because I've only ever dabbled on the side.

It's like my one discretionary play trading venture left because everything else is purely systematic. And I got hurt by Luna, which was a coin that went from, I can't remember, it was like a $100 and it went to 0.01 of a cent or something overnight. So there's risks with the coins.

There's a lot of corruption around and then there's exchanges that have like FTX that have gone completely underwater. Talk to me a bit about how you do your risk management and where are we at in the industry now in terms of these risks?

Pavel Kýček 16:36: Yeah, no, this is very good, very good question, Tristan. Well, firstly, these are two risks. One risk is, let's say, left tail risk of single coin position.

Another one is exchange risk. So firstly, let's talk about the left tail risk, which is significantly higher. And that's true compared to TradFi trading if you are not playing with very high leverage.

So to left tail risk, honestly almost forgot it because the way how we are approaching this risk is really reducing the risk a lot. Of course, the best approach is reducing your single position to a few percentages, ideally without any leverage. And that's basically what we are trading.

The best solution is really to trade across 15 to 20 different models. Each model should be trading up to 10 to 20 open positions in one time. And then you are ending up with a position anywhere between 0.5% to 3% based on your targeted volatility and based on the exposure you want to get.

And that way, really, as long as you are able to keep some management constraints, some liquidity constraints, you are pretty well covered. Honestly, as we are talking about the left tail risk, there is also the right tail advantage, which we are trying to catch through momentum models. So the coin really has two sides, which we are trying to take advantage of.

And the approach how to do it is really to spread spread the allocation across dozens, dozens of small positions in one time and then, of course, from time to time you are getting losses in dozens of percents, but again, on one position, it's making 0.5%, 0.3% of your account. So it's not a big problem for someone who is able to trade algorithmically fully automatically. The risk can be very well managed.

But of course, diversification is the way to go. And regarding exchange risk, well, there are solutions based on how big your account is. The simplest one is to spread your account across more exchanges.

For example, we are offering the solution to four exchanges. So again, you are able to cover it very simply then. Another like add on you can add to just spread the capital across exchanges is that you go with a little bit more volatile solution.

So you are keeping on exchanges just, let's say, 50% of the money, which should be for trade in crypto and then you are spreading across, for example, four exchanges so you are holding the risk of 12.5% per account which is already pretty good because if you are targeting some drawdown anywhere between 15% to 20%, for example, based on the portfolio, you are able to get a performance of 50% to 150% based on the year.

So this is the risk, which I think is already well paid. So this is for a retail trader or an individual.

And then institutional traders, they do use a solution which is called off exchange And they are using basically some middlemen, which is keeping the position for themselves. And they are just making on a daily basis the calculation of profit and profit losses, they don't need to hold the money on exchanges, which is solution many funds are using these days.

Tristan Pollock 20:31: Yeah, that's excellent. I presume also given the nature of your trading, you're not really trading the smaller, meme stock coins or what Well, it depends.

Pavel Kýček 20:45: Honestly, we are even trading shitcoins. We want to trade them, especially to the short side, because these shitcoins really tend to fall a lot during some bear market regimes. And for example, in this quarter, we have made anywhere between 15% to 25% just thanks to shorting these meme coins through our momentum short strategies.

So it depends because of course, again, higher left tail risk, but on the other hand, higher right tail advantage of making more money from a single trade. So then it's more about targeting some volatility, targeting the level of risk you want to get out of your solution. But we at Robuxio, we are trading the top 50 biggest coins, which we are ranking on a daily basis based on volume, market depth, liquidity.

And that's basically our tradable universe for today. And we are ranking every day.

Tristan Pollock 21:53: So when you say top 50, and we talk about the meme coins, are they the bottom of that top 50 or they're they're separate to the 50. They're like some really low down the list. Well, for example, these days on crypto futures exchange on Binance,

Pavel Kýček 22:11: there are somewhere around 300, probably even more crypto futures contracts listed, and we are trading top 50 based on some volume characteristics, the biggest one. Of course, you gather coins like Trump coin and these really pure meme coins. But again, because we are not trying to catch a single position and making a lot of money with one position only, we don't care that much if we are trading really shitcoins or bigger coins.

We only care about the exposure to meme coins, to low value projects and high value projects and to be really spread across the whole tradable, our tradable universe properly. And then of course, our solution, what we are doing for some of our institutional clients is that some of them, they, for example, have in the contracts that they cannot trade really top 30, top 50, really the smallest coins, which we are trading.

And for them, we are in trading on the universe of top 20 coins because these are less volatile.

The left tail risk is smaller and we can provide a solution too. So it really depends. As a retail trader, I wouldn't be afraid of these small coins, especially if you want to trade mean reversion strategies and if you want to short too, because the edges to the short side of these small coins are really, really big and doesn't make sense not to exploit them.

As long as liquidity constraints are not a problem for you. And liquidity constraints, you don't have to be afraid of it as long as you are trading anywhere between few millions of capital.

Tristan Pollock 24:04: All right. Well, let's get into your models and your strategies and just starting with the data that we've already mentioned. You kind of need a different process for building strategies on a limited amount of historic data.

So what's your approach to trading assets that have very minimal history, but also in one sense, very biased history, I guess, through the likes of they're in an introductory phase, they've grown enormously. What's your approach to building? What's the first steps in your approach to building robust strategies with less data?

Pavel Kýček 24:44: Yeah. So for us, it's something to data because it's also interesting point. In crypto, have like seven, eight years of data, but to tell you through any data before 2020, you almost cannot use because it is the year when crypto futures started trading.

So that's one thing. So you really have just a few years of data. So how we are approaching it?

Firstly, we are building all our models based on price and volume. We are not using any crypto only related data because we really don't have enough of them to rely on them. Secondly, all the models you you are building or you would be building should be made based on some research on any other asset class.

We are trading momentum approaches. We are trading mean reversion approaches. All the models we are trading would be tradable somewhere else on stocks, on commodities usually.

And then, of course, you have to make some changes to crypto because crypto compared to any other asset class today is much more dynamic. It's trending much faster, but also the regime changes are much faster compared to stocks, compared to commodities. So our approach is to start always with idea for solution, which means that we know what kind of edge we want to catch.

Right? We want to trade, for example, momentum long. So we look at the research and personally, I'm also trading on other asset classes.

So I have some experience from from stocks and commodities. So firstly, you look at what's working on commodities and stocks depends on what type of momentum model you would want to trade. And you go through the research, you test it on commodities and stocks, you take the model to crypto and you make as few changes as possible.

The main change is almost always in exits because exits are really reflecting the dynamic dynamic characteristics of any asset class and this is how you are adjusting to the volatility of crypto in general. So this is one thing. Another one is that we are always using as little conditions as possible.

Usually, we are moving anywhere between two to four conditions per model, which is like four is really the maximum I would go with because the more conditions, the more complexity exponentially you are adding to the tradable model. And then, of course, we are running many, many different robustness tests, which we built in house.

It's not about like normal robustness tests you would run on stocks or commodities where you have the advantage of having dozens of years of history.

It's more about, for example, running your own daily data. For example, we are building in house our own daily data with offset daily closes to test the robustness based on the daily close, to test everything on lower timeframes for eight, twelve hours, to test that the model is stable across broad vary variety of parameters.

So there are many, many steps you have to follow, but the most important one is to take model which is tradable somewhere somewhere else as simple as possible, as robust as possible, and deploy to crypto with as few changes as possible.

But still, you cannot rely that this model won't stop working because as you correctly said, we are still in some immature phase in crypto and we know the only thing we know honestly is that crypto will be more mature over the long term, but we just don't know how it will evolve. Will it be more like stocks or will it be more like commodities or more like Forex? We don't know.

That's why my approach is to really trade broad portfolios or whether portfolios of long short momentum mean reversion together because that's the only way how I can be really prepared for any market environment which can happen in the future.

And then, of course, your work and our work as traders is to have proper benchmarks to all the models in the portfolio you are running and comparing the performance of the model to the benchmark which you should have again built in house because there are no proper benchmarks for now.

Tristan Pollock 30:00: I've been speaking to some shorter term futures traders on the pod recently, Pavel, as. And so there's a bit of a difference between the way futures traders or even currency traders generally approach their trading and portfolio building compared to a stock trader. And I bet you know where I'm going with this.

Stock traders, because you do some stock trading as well, we tend to build models that run over the whole portfolio, the whole universe of stocks so that we're taking a portfolio based approach. We want to build alphas that would apply to any of the selected stock that we happen to trade that day. We're not building strategies for a given stock.

And the futures traders, the shorter term ones that I'm speaking to recently specifically, they are of course generally making models that are tuned to the particular market that they're trading because they are quite different markets. Could be trading bond futures, equity futures, commodity futures and so on. Where do you sit with crypto?

Are you making models that you want to be robust across the universe of say, the top 50 or Yeah. Are you making, individual models per per No. No.

Always always portfolio approach.

Pavel Kýček 31:23: Why? Because the hindsight bias in crypto is huge. Honestly, wouldn't believe there are even institutional traders who are building their models on top of few coins, which were performing very well in the past.

You know, for example, Solana. Solana was making hundreds and thousands of percents in a few years. They are putting some momentum model on top of that.

And they have to believe that such a model would be able to perform the same way. Honestly, I wouldn't believe that it's even possible, but yeah, the whole industry is still pretty immature. So how we are approaching it is, as you said, basically stock portfolio approach.

All our models are built that way that they are tradable across the whole tradable universe. Of course, there are models which are more and again, that's something you can even observe in stocks. There are models like long term momentum aka trend following is doing better on bigger projects.

That's also something you can observe in stocks too, that bigger stocks tend to really create long term trends. This is something we can already see in crypto too. On the other hand, we are not betting on this behavior because again, you can expect it based on research on other asset classes, but you cannot 100% bet on behavior to be continuing over the long term.

But short answer is always portfolio approach, always without any hindsight bias.

Tristan Pollock 33:07: I think you've answered a question I was going to ask about specific things that might work in crypto, that don't necessarily work elsewhere. But the whole approach to trading this in environment with these new markets is to take the principles that you know work elsewhere. And so in a sense, are trying to match that up and taking this logical approach.

And, what I like about that is, it shows where traders make their money using their, their logic, using their mind in a way above and beyond just running code and running the numbers. You need to have some entrepreneurial sense about you. Always make this point to build something robust and so you've taken that thinking and, it makes a lot of sense to approach crypto in that way.

And I know you're having a lot of success with it. So are there particular, let's drill further into your strategies. Are there particular strategies that you'd like?

Like what's your spread look like? You've said, you mentioned about 15 to 20 models, break that down for us kind of long term, short term, long short, mean reversion trend. What does the spread look like for you?

Yeah.

Pavel Kýček 34:33: So first, let me tell tell you one thing. Big. I really like trend following long term momentum.

I think it's really one of the most robust approaches over all the asset classes. On the other hand, it's also the most volatile approach, especially if you don't have the benefit, the advantage of really diversify of diversifying across different asset classes. And because we are running our solution on top of crypto only, we need to get the diversification differently, which means that we are running momentum and mean reversion.

And by momentum, I would like to stress a little bit more the difference between short term momentum and long term momentum. Because short term momentum, for example, on stocks, it's something which is almost not working, let's say, term breakouts to the long side. If you really dig deeper in research, you can see that big or short term breakouts tend to be mean reverted on stocks.

It was working in the past, it's not working anymore. That's why your good point is not only on focusing on what's working today in TradFi, in traditional assets, but also what was working because that's also where we are in crypto right now. So short term momentum and by short term, if you would be trading on daily data, short term is anywhere between one to three days of holding is very strong approach, especially to the long side.

Then momentum, like long term momentum aka trend following is making a lot of money, but with very, very short time windows or periods because crypto tends to really overreact during very short time and then it tends to be really sideways or falling back again. So we are taking advantage of momentum, mean reversion, long short. Of course, long side is very similar or this characteristics is very similar to stocks.

Long side is much much stronger compared to trading through short side. So being retail trader only and if I wouldn't have fully automated solution, I would go with momentum long and mean reversion long in broader portfolio because we have the advantage of full automation. We are going also with the short approaches.

These have much smaller edges because crypto in general is mean reversion long asset class, which means that already dips tend to tend to react to the long side most of the time. But because we are approaching portfolio approach, all weather approach, we want to have this exposure not only for making money during bear markets or bear market regimes, but also for hedging our long positions

because the way how we are approaching trading is to spread the allocation based on the current market regime but we always want to be hedged if there is some left tail environment, something happens, the whole crypto market falls by 20%, 30%, 40% even during the day, you have to be sure that you are managing this type of risk. And in crypto, one of the worst way how to manage risks is through stop losses, hard stops.

Because crypto is really very volatile, especially intraday, and it's making many false breakouts and many fakeouts in general. So there are smarter ways how to how to approach this risk. And based on my experience, the smarter way is really to have models.

And now we are getting more to the models which we are trading to have models which are trying to catch the edge and models which are hedging your portfolio. Let me give an example. For example, I'm talking about us having 20 plus different models which are running in production.

Of course, we are trading momentum and mean reversion only, so this is not about finding dozens of different edges, but it's about how we are approaching each model. Some of the models are, for example, more aggressive in a sense that they are getting into the market anytime there is some spike in volatility. Anytime the volatility is moving, these models are getting to the market.

Why? Because they are giving us some initial exposure to the market plus these are great hedges for momentum models to the other side. And then we have models which are really regime dependent and these are only let's say switched on anytime the market starts doing its thing, it's starting really moving to the long side a lot or to the short side a lot.

The difference between these models is usually in expectancy, The models which are built only to catch the biggest edge out of the market are definitely better in a sense that you would want to trade them by themselves only, but the other ones are there to build properly well built all weather portfolio.

Tristan Pollock 40:26: Yeah, we talked about this approach with PJ Sutherland on the show who has spent a significant amount of time building, mean reversion and short term breakout strategies specifically because of how they will hedge one another and synchronize with one another.

That So hedge that you're building in is not only significant and important, but it allows you to compound your returns better because you're not getting yourself into drawdown so you can keep making money.

Pavel Kýček 41:06: Plus you have made one very good point. And I think that's something even momentum traders should be thinking about. And that's the combination of momentum and version.

Let me give an example. We all know that for most momentum models, profit targets are not the best way how to exit a trade. Usually you should let the market run, right?

But especially in crypto, want to let the market run because it can make hundreds of percent per position, especially on these smaller points. On the other hand, the bigger the open position and not on one coin, you can do it through volatility targeting,

but the bigger the position to, for example, momentum long approaches in the market as the market is approaching some period when it's really growing a lot, you are having big open profits and also potentially big drawdowns anytime the market starts reverting a lot.

How you can reduce these open profits or the modern open profits, the open exposure to the long side is by using very short term mean reversion short strategies, which are using the position anywhere between few hours up to one day.

Usually we personally are never holding any mean reversion short longer than twenty four hours, because the left tail risk is generally getting much bigger. But thanks to this combination of momentum long and mean reversion short, you are able to push the long exposure lower and you are able to reduce this left tail risk of going very down out of your open profits and being exposed to other approach anytime.

Tristan Pollock 43:00: Yeah,

Pavel Kýček 43:01: I love it. And the same to the short side actually. For the short side, I believe it's even more important because as I said, crypto market is really mean reverting to the long side, but you want to have this momentum short exposure because for example, in a bear market of 2022, you were able to make with a bear or momentum short models performance profits anywhere between 40 to 80%.

So pretty nice one and it's really stabilizing the portfolio. On the other hand, most of the time momentum short is eating your profits. Again, to reduce this disadvantage through mean reversion long, because the stronger the momentum to the short side, the higher the probability that the market would mean revert back to the long side.

And you can take advantage of it and you can deploy different mean reversion long models and they are really working very nicely together in the all weather portfolio.

Tristan Pollock 44:06: For the single strategies before you start to combine them together in your portfolio, Pavel, can you give us an idea of maybe what the metrics look like on a single strategy in terms of say return and drawdown because I'm positive that once you combine those 15 to 20 over 50 markets, the portfolio numbers look substantially different. So it's good for the traders to get a feel for what that first step looks like and what to expect on a So single

Pavel Kýček 44:42: again, it's very important to differentiate between different edges, momentum long with short term momentum you are able to make on average 60 to 80% per year with a drawdown of 20%. With long term momentum on crypto, the average holding time is anywhere between seven to ten days on average. We are talking with slightly higher performance, but you are paying with higher drawdown.

So if you would be able to go through 40% drawdown, can make up to 100%, 120%, even 150% of profits. Maybe we haven't mentioned one thing at the start. I think that's very important and it's why we are able to make these profits.

It's because the volatility of the market. If we, for example, compare the volatility of stocks of S&P 500 and compare it to the volatility of our tradable universe of top 50 biggest coins, we are talking about seven to 10x higher volatility of this crypto tradeable universe. So that's the magic, I just wanted to really stress it that it's not about finding some magical receipt of how to approach crypto and make a lot of money.

That's really about how to manage the left tail risk while being exposed to this tremendous volatility. And that's why this all weather approach. But sorry, just to get back to your question.

So momentum long, we are talking about 60 to 80% with a drawdown of 20%. We are talking about short term momentum, long term momentum, everything even higher. Momentum to the short side, that's about 20% average annual performance with a drawdown of 20 to 30%.

It's definitely much, much smaller. Mean reversion long, that's interesting one because there are two approaches to mean reversion in general. One is really the one that you are buying something which was falling a lot and you are betting that because it really fell a lot, it tends to mean revert oversimplified.

Another one is and the one I like a lot. This is combining momentum and mean reversion together because we do have both approaches. One is really pure mean reversion type of trading.

Another one is the combination of momentum and mean reversion. And this one in crypto is very interesting because if you are able to measure momentum and then you are entering on some type of mean reversion characteristics of the market intraday or during several day correction, you are able to get the best of both worlds, which means a higher win rate which you are getting through mean reversion and also the exposure to the momentum effect

which you are getting through momentum strategies. Of course, it also has some disadvantages, because you don't have to get into all the strong movements, which are not making the correction.

So everything has its advantage and disadvantage, that's why we want to build a proper portfolio. Mean Reversion Long, we are talking about few dozens of percent on a yearly basis with twenty-thirty percent drawdown. To the short side, momentum short, if you are able to trade it properly with proper diversification is pretty strong edge in crypto, because I feel that not many are taking this side of the market that often.

So this is something worth exploring, but you have to have really automated solution for this one because you have to manage the left tail risks, which are much bigger with mean reversion shorts, especially on crypto. And whenever you are able to manage them, then I definitely recommend going over it.

Tristan Pollock 49:07: How do you combine all of these strategies into the portfolio? Do you have some set allocations such as, like I must have 25% in each of four sectors, 25 long mean reversion, 25% short mean reversion, 25% long trend and so on. Do you break it up like that and then try and keep it in those sectors so that they're always available to offset one another or, as do you have some other rule?

Pavel Kýček 49:37: Yeah, oversimplify that's what we are doing. Basically, we are saying what's the exposure we want to get from different approaches And that's something we are keeping. But there are other constraints you have to keep in the portfolio, especially maximum position per coin, maximum.

And then it's really more about the exposure you want to get from your portfolio, because it would be too simple if I would say that we are doing it just this way, because what we are doing for our institutional clients, for example, is that we are offering portfolios with different kind of exposure. Let me give an example. We are trading for a fund.

They have strong momentum long exposure and with our solution they want to get a higher exposure to mean reversion approaches and momentum short approaches. And that's something we are able to do. We are doing it through building portfolios based on the models which are running in production, which means that we are not starting building models from scratch.

These models are running live for a few years. We are taking them and based on the target exposure, target volatility, we are building the portfolios. So in general, for myself, I would build a portfolio with the exposure something like 60% momentum, 40% mean reversion, 60/40 long-short, more or less oversimplified.

But the truth is that right now we are running over 20 different portfolios for our clients with different level of volatility, with different level exposure. Understood.

Tristan Pollock 51:26: I would imagine that most of the models you're building, Pavel, you build them based on those that cause a logic that you talked about so that they're robust so that you can let them run. Do you do that or do you have a process to actually retire strategies and substitute new strategies in fairly frequently?

Pavel Kýček 51:49: Yeah, so definitely very frequently, but we have this process. First of all, you have to have the proper benchmarks for the models, which means that the truth is that we are running over 20 different models live but we have dozens of different models which we have built and which are very similar to what we are running live.

And of these similar models, we are building our own indexes or benchmarks, which we are then comparing on weekly and monthly basis to the models which we are running live.

That's the only way how I the only way what I found how to do it very properly. And then of course you have to understand the models. I think that's very important piece of the puzzle because let me give you an example.

You know that you are trading a momentum short model in the portfolio. The market is going through two years of bull market and another one year of sideways market. In such conditions, what should be the output of your momentum short model?

Well, I wouldn't expect that this type of model would be making money, right? Especially if it would be a model which is made also to hedge our momentum long positions. This model for sure would lose money for these three years and it's something I would perfectly expect.

So I think how we are approaching it is that firstly we know what we are trading and we know in which market phases the models should be making money and in which phases they should be losing or they should stay completely out of the market. With this level of understanding, you can already say most of the time that the model is doing good job, bad job, it's some exception or something is going wrong with your model.

But we are adding the other layer of benchmarking all the models we are running in portfolio and thanks to them, we are getting some soft big picture and soft comparison, but also hard data comparison, which we are then putting together.

And based on this, we can make the decision to, for example, discontinue some model or take some other model and put it into the portfolio instead of the one we are trading. But in general, these changes are not made like on a monthly basis, not even on a quarterly basis.

We are really talking about very minor changes on a yearly basis because we also need to build something which is giving the same exposure which our clients, especially the institutional ones are expecting.

Tristan Pollock 54:55: Yeah, that's what I meant. And that's what I thought. So tell us a little bit about your tech stack because you've talked about the importance of the data and getting the right data, cleaning the data, but then executing potentially executing for different clients on different exchanges or across multiple exchanges.

There's some really significant technological requirements there for your order management and your data management. Yeah. Tell us a bit about the tech environment in which or the tech environment you've built out to handle all of that.

Pavel Kýček 55:36: Honestly, what's good to say is that Robuxio is not about me. It's about these days about 10 people which are working on the solution on a daily basis. That's that's one thing.

Another one to your question, it's really heavily software related company because one thing is building the portfolio, building the models. Honestly, crypto has the advantage that if you take simple, robust model and you don't expect 45 degrees equity curve and you are able to trade momentum long, let's say only, you will make probably money over the long term because this edge is pretty strong. Of course, the volatility will be huge.

You will have years of sideways account, but you will get it there if you have the patience. But if you want to trade all weather approach with dozens of models, which means that you have up to 80 open positions in one time, another dozens of pending orders, you have to run your risk management and everything. It's heavily, heavily, really software related.

For example, right now we have a team of six software developers, cloud architects and software architects who are working on the solution on a daily basis. And it's really about having the solution, for example, how it is working in house in Robuxio. Imagine that you would start trading with us.

There is an automated process, which is building the portfolio in Amazon in the cloud and it's keeping the benchmark of the portfolio you are running. Then based on some frequency, let's say on twenty four hour frequency, we are making the decision on what trades to do, what to trade, what orders to send, what orders to cancel and so on. Basically, you are running your portfolio, right?

But this is where the work just starts because we are sending all the orders through API across many exchanges, which means that you really have to maintain the proper API bridge which is always changing And then you have to be 100% sure that you always know what's going on all the clients sub accounts.

And it's built that way that we are basically online checking what's going on all the client sub accounts and we are comparing it to all the benchmarks, their portfolio benchmarks we are keeping at Robuxio. And there are automated steps which are making proper changes if needed and everything.

So it's all about the robustness because the idea above which we started building the company is to bring traditional approach to trading, momentum mean reversion to crypto, take big all weather portfolio approach and trade as robust as possible. And by robust trading, especially on crypto, which is a market which is running 20 fourseven, the trading infrastructure is really something which is the main focus.

these days most of our resources are really going to the software development.

Of course, we are making our research on the daily basis, we are also onboarding another quant, so we are also having many trading related questions and topics, but the infrastructure is something which is pretty tough to build.

Tristan Pollock 59:26: I get it really important. Within So that infrastructure, how important is the execution in terms of minimizing slippage and that kind of thing? Is that is that really a problem yet for you guys?

Or do you have to actually create specific order execution algorithms, for example?

Pavel Kýček 59:52: It starts being a problem as we are growing, but we have built the solution. And as I told you before, one of the forms of robustness testing for us is to building our own daily data with different type of offsets. For example, we can build any number of daily candles based on few minutes offsets.

So this is something we are taking advantage of because of course you can run normal approach with how in TradFi people are dealing with liquidity constraints, which is TWAP-ing or VWAP-ing and these type of algorithms but we are approaching it slightly differently.

What we are doing is that anytime we see the client account is too big to handle like normal twenty four hour official daily data frequency, we are starting running these offsets which are offsetting the data on any number of offsets. So then we are spreading the liquidity across twenty four hour window.

The trading cost and slippage is not a problem anymore, plus we are adding different layer of diversification. As you can see, it is also giving different layer of difficulty through whole trading back end. So that's why we are so heavily in the technological part of our solution.

Plus what's important from the trading perspective and I think that's something many people are not doing is to focus on expectancy. Because me personally, I'm big believer that expectancy as long as you are building your models with idea first approach and you are really focusing on building logical models which should be surviving in the markets, then expectancy is the way how to approach robustness of the solution.

Because if you know that the expectancy on crypto of the whole portfolio is anywhere between 1.5% to 2.5%, then you know that small slippage, of course, it's eating the profit, but it's not ruining the model.

On the other hand, if you are digging in five minutes timeframe, the expectancy is usually very low. And as the edge will start decreasing and it always will, That's unfortunately the behavior of trading edge that it's getting smaller as it's exploited by more players. You really don't have enough space to make enough money.

That's why I'm a big believer with starting with as high expectancy as possible and see how it is working in the market over the long term. Yeah.

Tristan Pollock 1:02:56: Let's start to wrap up Pavel, but what about your just review process? Do you kind of review your strategies on a monthly, quarterly, annual basis? Or is it just when they start to decay in their performance?

What would you do to potentially review and update a strategy?

Pavel Kýček 1:03:20: Yeah, because I'm really enjoying building strategies in general, I'm on all the strategies almost all the time, so it's hard to say for me was the reviewing period. But we are always comparing our strategies on weekly and monthly basis or the weekly is very low one for daily data. But I just want to understand also the models, how they are performing under different circumstances and in different periods.

And if you really want to understand your trading model, I believe you also have to follow it very thoroughly even on the daily basis to see how it is interacting with different market behavior. So this is one thing. Another one is that we are benchmarking our models on monthly, quarterly and yearly basis.

And yeah, then of course I'm trying to understand if the model is doing what it is supposed to do. So again, was it made to catch momentum long? If yes, is it making money during the periods when the market is in momentum long characteristics?

If yes, I'm pretty okay with it and I'm continuing trading this this model further And if not, I'm starting digging deeper. What's important to say is that compared to TradFi, compared to stocks and commodities, you will see that the edge will decay quicker on crypto.

Because these days really, for example, we have seen in last six to nine months, a lot of new institutional investors and traders are rushing into crypto because they can see the potential of crypto trading.

And that's why I would expect that these edges will be getting smaller and smaller over the long term and that's why the metrics will be also worse. So the proper way how to do it is probably not to focus on metrics only, but really we are getting back to understanding to what you are trading and put in everything to proper picture.

Tristan Pollock 1:05:35: We've talked a lot about robustness, I think, and I wanted to call it out as a specific point to get into it specifically, but I think you've just peppered this whole show with different principles about how you test robustness. So because it is important and specifically important for crypto with a short data history, I do just want to call it out again and maybe summarize some of those points that you've made and you can add them.

Things that come to mind, of course, that you've got this causal logic based approach to building the strategies in the first place.

So you're not really into machine learning or data mining out of that short period of crypto data. You're working on price and volume. You're not looking at the other crypto related data that's quite minimal so that you can apply principles that we've learned from TradFi so that we're doing something that we know should work.

You've talked about manipulating the data, making different different end of day data. And I think I understood what you meant that you're basically changing when the bars start and end essentially. So you're making alternative data sets and you can test it on there.

You're also building strategies for a portfolio of crypto of coins so that it's not just super overfit to one particular market. What have I missed? So some of the other things that you've mentioned already or that come to mind, like what what's some of those key principles for you to really summarize and wrap it all up around building a robust strategies in the crypto space?

Pavel Kýček 1:07:27: Well, even our name is partially including the word robust. So I'm building, we are building everything around robustness. And you mentioned very well that it's you have to start with the models, then the robustness have to really be pushed to the whole portfolio.

And the most important thing, one of the most important things is really not to rely on the metrics in the past, but building models which you think will survive in the live market. This is what robustness is about to survive in the live market because in crypto, crypto is really not an investment asset class for now. Crypto is really asset class which is the best for trading because of volatility and inefficiencies.

So why we are so focused on robustness? Because the volatility and inefficiencies is in the market, but there is also huge left tail risk. How to avoid this left tail risk?

You have to have robust models, robust portfolio, robust trading engine and you have to be able to reduce the left tail risk and be exposed to the volatility and inefficiencies and to be exposed on a daily basis on the market which is running 20 fourseven. It's not only about robust models, but it's also about having the robust infrastructure which is able to trade really in crypto.

I cannot imagine the proper way how to trade it without some form of automatization.

Tristan Pollock 1:09:09: Pavel, to take us out, I'd like to know whether you have kind of a view on the whole economics of crypto, you could have zero interest in the, the more philosophical part of crypto and that would be fine because you are there to trade the markets and make those markets more efficient in the process and extract profits for your clients. But do you have any opinions? Are you like a Bitcoin maximalist?

Do you think that crypto is the future? Do you think that 99% of coins are gonna die? Do you have any sort of bigger picture philosophical and economic

Pavel Kýček 1:09:54: viewpoints on crypto and where we're going? Yeah. So first I have to say I'm very bad in making predictions, but I'm having long term bet in Bitcoin.

That's the only crypto I'm holding over the long term. That's something I can say publicly. Everything else, what I believe in crypto is really in a phase which is very comparable to tech stocks before 2000.

There were Amazons, there were many big companies, but there were also many projects which didn't deliver any value. They didn't deliver any promises. And this is exactly what's going on in crypto.

Crypto won't go away. You can see that in Asia, in US, in Europe, everywhere, there is the regulation being built these days, because everyone can see there is a huge opportunity whenever there is new asset class which is being bought, which is right now in crypto. But we simply don't know how crypto will evolve.

I believe that it could potentially be also a market which would be working as something as stock market for like IPOs of smaller projects, for example, that's something how it could evolve. We can see that there are many projects which are really delivering what they promised, but still what you said is true. I would say that not 99%.

These days, I would say 99.99% of all crypto coins will definitely die, which doesn't mean that during their process of dying, they won't make trends of hundreds of percent. And that's why we are here. And that's something we are trying to catch and make money from.

Tristan Pollock 1:11:50: I had the thought today thinking about us chatting and one really positive attribute of what crypto has brought to the world, think. I'm economist trained. I've got that economics background.

And so I've always been fascinated by monetary economics and how money works and how money is created. And actually, I'll have to post a link. There's a guy who explains money creation better than anyone I've heard.

I think there's a book called Prince of the Yen. People will know it. And there's a video of his where he explains money creation so brilliantly.

But anyway, so I'm interested in all of that. And I do think one thing that's really interesting with crypto is it's getting a lot of people thinking about money creation. And of course now, with the inflation that we've had, it's even more of interest and people are therefore more interested in economic history and just how long the fiat currency experiment that we're in has been going, which is not very long at all in human history.

So that I find really fascinating. So we really don't know what's going to happen next, but people are more aware than ever of the nature of money in the economy and how it works, how it's manipulated and controlled as well. And sometimes you're good, sometimes you're bad.

So I find that really interesting.

Pavel Kýček 1:13:34: Yeah, completely agree. You can see how people, especially Bitcoiners, the ones who are holding Bitcoin, they are having very, very good understanding of how money is working. But even people in crypto in general, if it is advantage or disadvantage for the whole system, I'm not talking about the individuals.

I also have my own belief and this is not exactly with the system and how it is being built, especially with the failed system. But there is the question because the whole system failed system is built based on the trust and the trust is starting being a little bit bitten through these people. So then the question is if it can help or not help over the long term.

You know what I mean? I like that people are starting understanding what's going on in the economy because some economists do. So I like it a lot.

On the other hand, because everything is built based on trust, who knows how it all ends up.

Tristan Pollock 1:14:45: Correct. Correct. Well, yeah, we might be living through some tough times.

Maybe we have to live through some kind of regime shift, right? Maybe. Let's see.

Hey, That's Pavel, big topic for another podcast. Yeah, that's above my pay grade. I really appreciate it.

Thanks so much. Obviously, you've got the managed accounts that you run there. There's so much going on.

So let people know, your website and how to get in contact with Robuxio.

Pavel Kýček 1:15:15: Yes. So the best way how to approach us is to go to robuxio.com and go over the website. And I believe there is pretty good explanation of how we are operating, what we are doing.

And if you are an institutional investor or individual trader who is interested in allocation crypto, just let us know. There are links. Awesome.

What's your X handle again? Pkycek. Got it.

Tristan Pollock 1:15:49: I'll start Also it

Pavel Kýček 1:15:50: on LinkedIn.

Tristan Pollock 1:15:51: Yeah. I'll put some links in the show notes. Well, once again, Pavel.

Really appreciate it, mate. And, look forward to chatting again soon. Keep in touch.

Thank you, Tristan. It was great to be here.

Show intro 1:16:03: 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 adviser. 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.