Top crypto expert reveals 220% return trading strategy

Transcript

Pavel Kýček 0:00: Volatility, that's the main driver of potential profits. What we have to know is that long term trends are about value and the growing expected value and short term trends can have many triggers. For you who don't know what mean reversion is, that's basically an approach that is trading against short term market overreactions.

You always or almost always get better results without having this fixed stop. Why? Because of intraday mean reversion characteristics of the market.

If you want to start with algorithmic trading, don't try to automate everything.

Richard Moglen 0:45: Okay, welcome back everybody to the TraderLion conference. I'm super excited to be introducing our next presenter Pavel Kýček. He is a fully systematic trader with over eighteen years of experience and his specialty is algorithmic strategies in stocks and cryptocurrencies.

This is a presentation I've been really looking forward to personally. Pavel, thank you so much for being a part of this and looking forward to dive right in and asking you a ton of questions as well. I'm gonna really enjoy this one.

Pavel Kýček 1:13: Hi, Richard. Thank you. Thank you for having me here and yeah, let's jump on it.

Let's do it. Okay, so let's go to algorithmic trading and crypto and how to put everything together because I think most people are making many mistakes regarding to algorithmic trading and regarding to crypto investing and crypto trading in general. So let's cover it.

Firstly, some disclaimer, of course, some of those results are back tested one. Back tested results are never the same as the real ones and keep safe and always stick to your risk management. So my first message is don't invest in crypto.

I know that it's pretty strange because we are talking about crypto trading, crypto investing here. But really what I'm always trying to say is that investing, and by investing, I mean long term allocation of capital into different crypto projects, is usually the worst way how to get allocation to crypto. Why?

Because if you compare, for example, S&P 500, the stock index that everyone knows to S&P Crypto since 2018, you can see that the performance is more or less the same, but with huge, huge volatility. Of course, if we would start here, the performance would be higher, but the main message is that crypto is not an investment asset and it has its reason that we will cover a little bit later.

One of the main reasons why not to invest into crypto is that most of you, and I mean, really most of you wouldn't be able to withstand drawdowns of minus 85% or minus 80% or another minus 80%, it's just huge.

And by my experience, I know that the drawdown of minus 40, everything between minus 20 and minus 40 is usually not able to go through during live investing or live trading because going over these results on the chart is totally different than to live through these movements. So really don't invest, do it slightly different. Why not invest in?

I really like this table because if you want to compare some assets, some trading approaches, some investment approaches, you have to have some basic comparison that you can use across all the asset classes, all these investment approaches. What I often use for very first comparison is basically a risk reward table, which is giving us the ratio between annual return and drawdown.

And if you divide your annual return of the asset or the strategy by the maximum drawdown, you get some ratio that you can start calculating with and you can start comparing this ratio between each other.

Here, if you compare S&P 500, just buying and holding S&P 500 since 2018, that's the time when we are starting having good data for crypto too. To buying and holding S&P Crypto, you get much, much, much worse return to drawdown ratio compared to S&P 500. Why?

Because of this huge, huge maximum drawdown. In fact, if you would like to get to the same ratio on crypto to S&P 500, the annual return would have to be at least 26 or 27% average, which I'm really not sure that crypto will be able to deliver such a huge return just by buying and holding over the long term. Plus what's even more important is that drawdown or volatility is just one of many potential risks.

In fact, you should adhere many, many more risks that crypto is still full of. So really this is let's say the basic information and we will start building on this table later on. Why we are in crypto?

Why we trade crypto for our clients, for institution and retail clients is crypto volatility, not long term friendliness. It's super important because compare the blue line here, which is basically ATR 14 of S&P Crypto to S&P 500. Again, I like these comparisons because it is really giving eye opening view on crypto in general for those that are not that familiar in crypto.

You can see that the difference between S&P 500 and S&P Crypto is between 300% to up to 1500%. So the volatility on average over the last five years is about 800% or 9x and this is the reason why we all who like trading should be in crypto because volatility is the key, that's the main driver of potential profits. So this is something that is extremely important to understand.

And it is something that we should think about if we are building our strategies, if we are building our projects. Now, what's volatility? Just for all of us to know, volatility basically is unit of movement.

The higher the volatility, the better for us traders, of course. And what's even more important is to understand the basic characteristics of volatility and these are two main ones. The first one is that increased volatility tends to remain increased.

You can see it on this chart on the Y axis, you can see the day to day volatility and here you can see the years, basically days. And you always have periods when the volatility is much higher and those periods are always followed by periods when the volatility is much lower and it's always changing. The periods of high volatility are followed by the periods of low volatility.

In fact, this is the only 100% market cycle that you can bet on that after periods of low volatility, there will be always periods of high volatility and vice versa. And the second fundamental characteristics of volatility is that increased volatility tends to create at least short term trends. This is something that we can bet on over the long term.

What it means that usually if we have some initial point of view and the market, there is some news let's say on the market and there is some increased volatility. This increased volatility with some probability of course usually pushes the price into some following volatility and following trends.

And based on these two fundamentals, you can start building your trading models because your trading models always have to consist some fundamental condition that is based on the main characteristics of markets in general.

So volatility is the key everywhere, especially in crypto. Now, another thing is that we should know the main reasons or the main differences between long term trends and short term trends. Because long term trends, very simplified, are about value.

It's about expecting higher value in the future, especially if we are talking about stocks, but these short term trends can have many triggers, which is key. It can be health behavior, this can be manipulations or even different pumps and dump schemes, especially in crypto. At the end of the day, doesn't matter.

What we have to know is that long term trends are about value and the growing expected value and short term trends can have many triggers and you can see it on this chart again, the blue line S&P 500, you can see perfectly nice long term trend, Compare it to S&P Crypto, you can see many long short term trends to the long side, to the short side, to the long side, to the short side, then some flat and then to the short side again.

But these are definitely not long term, definitely not in crypto, because there are too many projects that are not delivering value. They are trying to find some value, but they are not delivering it and that's why these long term trends are super volatile and they are not that smooth.

We should bet on short term trends more than on long term ones. Now let's get to very first trading model that I really love because it is showing the strength of algorithmic trading on crypto and it is showing the inefficiency of crypto market in general.

This trading model is very simple, it's only about buying Bitcoin, right now we are talking only about Bitcoin, if it is closing above its moving average 50 and we close the position if Bitcoin is closing below its moving average 50.

Let's see that we have some chart, we have some moving average. And if we close above it, we get into positions and we stick with the positions as long as Bitcoin is above its moving average. And once it is closing below, we are getting out of this position.

And if you compare just buying and holding Bitcoin to this very, very simple trading model, you get the overperformance of 7x or six to 700%, which is huge. If you compare it to stocks for example, these very basic trending models would give you probably underperformance to just buying and holding because of efficiency of traditional assets.

The less the asset is efficient, the more these very basic trading models will deliver over performance compared to just buying and holding.

So it can be even your basic comparison among different kind of assets or asset classes. But it was only about Bitcoin. And I don't like this example from because of one reason and it's because of selection bias.

selection bias and hindsight bias is something you wouldn't even believe. How many even professional traders or capital allocators are suffering from?

They just choose Bitcoin, Ethereum, Solana, let's say the most known ones and they put some very basic trending models on them and they are showing these results from the past and they think that they would be able to deliver these profits to the future but it's huge nonsense and we will get there a little bit later but what's very important to know is that the more the asset is inefficient, the more the hindsight bias or selection bias is problem for traders and for real results of the trading model.

So let's get to one very simple trading model that most algorithmic traders know and it is rotational strategy. The idea is very simple and it again consists only of three conditions. One is context filters.

It means that we are filtering the market into phases when we want to enter into positions and to faces when we don't want to. So our context filter is the one that we have used on the example below, which means that we are entering to positions only if Bitcoin is above its moving average. If it is above its moving average 50, the context filter is switched on, and we can start trading.

And we only take trades on 10 coins that are the most trending ones based on rate of change 30. So we enter into these positions and we stick to these coins as long as these coins are in this top 10 positions by the trendiness. And once Bitcoin is below moving average, we are not trading and as long as it is below, we are just rotating the 10 most trending ones.

That simple, let's say that's very simple strategy that is even working on stocks, that's working on commodities, doesn't matter. But what's important is to check the performance. We are talking about, let's say 18,000% since 2019–2020.

Why? Because we are taking advantage of this huge volatility of smaller projects and this short term trendiness. There is no magic, this strategy will deliver the performance of 10 to 15% on stocks, and about 100% to 150% on crypto.

The main difference is the volatility of crypto in general. Because if we get back to this risk reward table and we calculate our ratio, we are getting to 1.6x. It means that a historical drawdown or the annual return is 1.6x higher than the maximum historical drawdown.

So we are getting onto something, it's super volatile solution, you wouldn't want to go through this phase for sure, but we are getting a model that is able to deliver annual return of 100 plus percent. Of course, percent drawdown is still too much, but it's still very simple one trading strategy. Now, how is it possible?

How is it possible? We have to break it here for a bit to even more understand, especially to those that are not trading crypto. How is it possible such huge returns?

It's all about the immature characteristics of crypto because once we started trading really professional crypto, I would say we started being professionals on immature to major cycles in general, because you could see these cycles everywhere over all these trading asset classes. Doesn't matter if it is stocks, commodities, but it's always the same.

Firstly, there is period which is about low liquidity, let's say the asset is pretty low liquid, which usually brings high volatility.

Why? Because you can have just pretty small trading sizes and these are already moving the market quite a lot. This high volatility usually brings big opportunities in terms of trends, short term ones at least and inefficiencies.

Inefficiencies on crypto that could be measured by mean reversion strategies are right now four to seven x higher, Joe, just for you to know. So these big opportunities, what they are doing? They are basically pulling new capital into the market and this new capital is pushing the liquidity higher and the higher the liquidity, the higher the efficiency of this asset class and the smaller the opportunities.

And this cycle is everywhere. You can see it on trend strategies on commodities in 1980s. At those times, trend strategies would be able to deliver 40–50% average yearly.

You could see it on tech stocks before the tech bubble for example. These were years that were very, comparable to current crypto market. So this is something that is always in the market and you just should or shouldn't based on your state and you should take the opportunity.

One more thing, you could even see that this cycle is already happening on Bitcoin. What these two charts are showing is the daily rate of change in ten days when the days were higher than 10% or 20%. You can see that in 2017, there were more than 50% of those days, 50% and in 2023, we are getting to under 20% which is still huge.

But these 20% plus days, these were almost 30% in 2017 compared to '23, we are getting to seven to 8%. So just to put it into perspective, this is one of the reasons why I was saying this that picking one crypto, Bitcoin, Ethereum or something and trade just on this doesn't make any sense because the volatility is just getting lower as the asset is getting higher and is more liquid.

But just very quick comparison, look at this chart and now compare it to S&P 500 and you can see that these 10 plus percent days were almost none in last six years and 20% plus.

These were only during COVID crisis in 2020 and probably don't forget the results in 2020, because all the models were delivering profits like above average. Why? Because of volatility, that's everything is about volatility.

So that's the reason why this trading model is able to deliver such a huge overperformance compared to just holding bitcoin. Now let's get to some real trend strategy on the whole crypto market, because I think it's even more important to show what algorithmic trading on crypto is in general.

We could pick out one of our strategies that we are trading for our clients, this could be trend catcher long and what it is doing is that it is trading based on moving average 20 all the tradable crypto futures that are filtered based on liquidity.

So we are not trading the smallest ones, but we are trading the more liquid ones, let's say top 50 to top 60 most liquid coins. And these are ranked on a daily basis. So there is again, no selection bias.

And don't forget that on this one is pretty high selection bias because we selected Bitcoin that was performing pretty well over last five years. And what we are doing here is that we are just buying coins that are above moving average 20.

And we are closing the positions if the coin is below moving average 20 and we are doing it only if the market is in the growing, let's say characteristics and it is only if bitcoin is above its moving average 50.

So that's the whole strategy, one entry condition, one exit condition, one market regime filter and some kind of ranking that is basically telling us that we want to trade the most trending coins only. That's it and again, look at the performance and again, why is it possible because of volatility and friendliness of the solution.

Because what's really important to understand if you are thinking about algorithmic trading and trading models in general is that these models can give only such a performance that the market is giving So that's not about magic, these models are definitely not magical, those are oversimplified because the more simple, the more robust the solution.

So they are really oversimplified, but you get there a little bit later. But they do deliver such profits because we are trading them on the most immature asset that we are able to trade right now. Now to key characteristics of a robust strategy, because it is something that every algorithmic trader should aim for.

First is logical basis. What it means is that it has to be made on some sound idea that is basically built into the market. We were talking about volatility, we were talking about that the increased volatility is basically pushing new trends.

Based on these ideas, you can build many trading models.

Another one is simplicity and that's really the key because every other condition that you give into the model is exponentially adding the difficulty of the strategy and also the probability of failure because it's super, super simple to build some strategy on past data but unfortunately, we are not trading history but we are only trading the futures data and that's why the more simple and the more logical the model is, the higher the probability of survival in the life markets.

And then there are two even more important points that most traders are not thinking about that much.

First is that every robust strategy is profitable only at certain market phase, that's really super important. What it means, if I have trend catcher long strategy, which is basically catching long trends, this strategy will deliver profits if the market is trending. So you can see here Bitcoin is trending, the strategy is delivering profit.

Here, Bitcoin is trending, the strategy is delivering huge profits, but if the market is going to the short side, your models will be flat or they will lose money. Usually, they will lose a bit of your equity and if you want to build one strategy that is able to trade and deliver profits in every market conditions, you will always end up with over optimized crap, always.

That's super important and the worst thing that most retail traders are doing is that they are trying to build on the past data some nice strategy that is able to deliver profits in every market regime and that's really nonsense.

And the last point that we covered somehow is that it will bring losses in phases that the strategy is not built for. Again, if the market will be super flat and we are trading trend strategy that needs volatility, the strategy will just lose money. That's it, that's the reality of algorithmic trading.

that's the reality of almost every trading approach, but most traders are just not thinking about it deep enough to go over it and start building portfolios, which is the key for algorithmic trading and we will get there. What is also important that especially if you are building strategies on daily data, you will have losing periods that can easily last for many months, even a year or two.

You just won't be able to reduce it by one strategy, you will have to start building portfolios.

To build very simple portfolio, let's just go over another strategy that we are trading. This strategy has just horrible equity curve, you definitely wouldn't want to trade by itself. You can see that it's basically flat for two years, then it made some money in one year and then it is again flat more than let's say one and a half year.

This strategy is the opposite of trend catcher long. So it is only trading to the short side, it is only trading if Bitcoin is below moving average 50 and it's only trading if the coin is closing below moving average 20 and getting out of the position if Bitcoin is, if the coin sorry is above moving average 20.

You can see that if the market was growing, for example here, this strategy was losing money, you can see it here, market was growing, strategy was losing, then the market was falling, the strategy was making money up to 80% and then the market started trading to the long side again and the strategy was giving some profit back.

You wouldn't want to trade this strategy by itself, but the value is very, very high and it's in putting this strategy into portfolio. Here you can see the very basic of portfolio, which is just two strategies together, trend catcher long and trend catcher short. You can see trend catcher long as the green line, trend catcher short as the red line, and the blue line is the whole portfolio.

Compare just trend catcher long, for example in this phase to the whole mini portfolio that we just built and you can clearly see that the stability is just thanks to these two models on very, very different level. And we are playing just with two models here. Normally, are trading ten, fifteen plus strategies.

So just to think about it again, just for some comparison, here you can see our very basic model that is trading on Bitcoin only, compare it to the TrendCatcher Long Short, the over performance is again 12 to 13 x or something like that. Pretty pretty high.

Richard Moglen 32:35: And sorry to jump in. Just a question about the combined strategy there. Is that equal allocation between the TrendCatcher long and TrendCatcher short.

So 50% in that, 50% in another, you're not kind of changing that as the models change. It's just kind of a flat equal That's great idea. In fact, in

Pavel Kýček 32:55: this particular example, we are getting allocation 100% to long and 100% to short. Why? Because we do have this regime filter that is basically switching off or on this long versus this short strategy.

So yeah, it is a little bit more aggressive, it is more like an example. But thanks to this regime filter, you can reuse the same capital for the whole strategy, which is super important and very strong way how to reuse the capital thanks to portfolio trading.

Richard Moglen 33:30: Gotcha,

Pavel Kýček 33:32: thanks. Yeah, you're welcome. So to portfolio trading and why it is key, just very quick sum up what's portfolio for algorithmic traders, for us algorithmic traders, it means trading two or more low correlated or uncorrelated strategies.

And it is very important because once you trade correlated strategies, the portfolio strength is basically none. Then it doesn't make any sense. So it's always about correlations of these strategies, because otherwise you can't take the advantage of portfolio trading in general.

And what's correlation just for you who don't know, it's basically a relationship between two variables. If the correlation is high, it means that the variables or our strategies, let's keep it simple, are moving the same way. If the correlation is negative, it means that if one is making money, another one is losing.

And if it is uncorrelated, it means more or less that one is making something and the other one is making complete like non correlated results. So why it is key? First one is something we already catched a bit and it's better capital allocation, because if the strategies are uncorrelated, in drawdowns, you can reuse the capital completely or part of it.

Thanks to this, the account is performing much better, because you are using the capital more often. So basically you are getting to higher usage of the capital, thanks to this higher allocation of capital. Then if the strategies are made properly, you are getting to higher stability of profits and you are reducing the risk of single strategy failure, which is something we didn't touch until now.

But it's also pretty important because every strategy can fail over the long term. Because even though you make it super robust, you make it based on the sound principles, based on the fundamental keys, key principles basically, your strategy can still fail. If you trade one strategy on your account only, you are basically, yeah, you are not in a good position.

If you trade 10 plus strategies, your clients or you on your equity curve won't even notice if you won't go deeper into the portfolio construction and you won't go strategy by strategy. So once you start building portfolio of 10 plus strategies, you are more of manager of more strategies in the whole portfolio.

And you can compare every single strategy to their benchmarks and you can see, yeah, this strategy is performing good, I can keep it, this strategy is underperforming more than a year, maybe there is something happening and you can make some steps and your training is much, much, much calmer compared to trying to trade one or two strategies when you never know when the strategy will fail and at the end of the day, probably almost every or close to every strategy will fail over the long run.

Now I mean decades, I'm not talking about months, but really decades of trading. So that's very, very important reason why to trade broad portfolios. Another one is that you are really lowering the risk of over optimization because if you trade single strategy, you are always pushed to get as much as possible of this single strategy.

It's just in us, know, we as traders want to perform well and if we want to perform well with one strategy, we are over optimizing it a bit, we are just making this very big mistake and over optimized strategy is always failing over the long term.

And another one is that if you trade more strategies like five, ten plus, you are even getting to smaller positions and the smaller the position, the more you are reducing the risk of single coin or single stock failure or some big drop in your positions. So you are basically minimizing many risks, but you don't have to reduce profits for those risks.

This is the reason why many traders are calling portfolio trading or algorithmic portfolio trading as the only holy grail of trading, because it's really possible to deliver nice performance and lowering possible risks. What's true is that if you understand this table, you very easily get ahead of 90%, maybe 95% of all retail traders. And this is correlation table of two strategies.

You can see it here. Here we have trend catcher long and trend catcher short, again trend catcher long and trend catcher short. If you look at a correlation in returns of trend catcher short and trend catcher long, it's minus 0.1.

It means that trend catcher long is delivering returns if trend catcher short is losing a bit and the opposite is true for trend catcher short. What's even more important is correlation in drawdowns. Because if you get to non correlated strategies or even negatively correlated strategies in drawdowns,

then it really means that you can push the performance higher while reducing the drawdowns and that's really the secret of long term profitable algorithmic trading.

Just these understanding these tables, you can see it here, we have some ROR of trend catcher here, of trend catcher short and the combined one and the maximum drawdown of trend catcher long and trend catcher short and you can see here that we really pushed the performance up while we were keeping almost the same maximum drawdown. Of course, here we are trading only two basic trading models.

If we put more strategies, what's happening more uncorrelated strategies, what's happening is that you can push the drawdown lower while keeping the performance the same and if you can really trade advanced portfolios of 10 or 50 strategies plus, you can push the performance higher while even pushing the max drawdown lower and that's something you never ever are able to make by trading single strategy, it's just not possible.

Let's push our portfolio slightly further and let's do it by adding mean reversion short strategies. For you who don't know what mean reversion is, that's basically an approach that is trading against short term market overreactions.

So let's say we have a market that is moving somehow during one day it makes some move that is two to three x higher than normal to one side, you basically enter against this move and speculate that the market will mean revert to some degree.

It doesn't have to mean revert fully back, but usually after these movements, there are some shorter term mean reverting characteristics of the market and you can take advantage of it and that's basically the second sound idea of how to build your trading strategy. First is Momentum one based on volatility and second one is mean reversion one based on market overreactions or also called inefficiencies.

You can see this model here, the equity curve it's pretty stable, it's also the characteristic of mean reversion strategies in general.

Why it is more stable than trend strategies is because we are in positions just few days, one to three days in general. And we don't need massive movements like in trend strategies, we just need one to two days of overreactions when we enter into and we are usually getting pretty high probability of profits between 55 to 70% compared to trend strategies that have the probability of profits of 35%,

but the average win rate or the average winning trade is much higher compared to the losing trade. So let's take this strategy and put it to our portfolio of trend catcher long and short and you can see the comparison here.

The green one was the original portfolio of trend catcher long and short only and the blue one is with our pump and dump or mean reversion short strategy. The main difference is that it is giving stability in the markets when there is like non movement or the market is not moving that much, the volatility is lower and this strategy is extracting money just thanks to these inefficiencies, not thanks to trends.

You can also see it here that is keeping the maximum drawdown slightly smaller, because trend strategies were basically giving back some of the profits.

This mean reversion strategy was making some money and that's why it was keeping the drawdown lower. Of course, we'll always have drawdowns even though we will trade 50 plus strategies in the portfolio. So this is with mean reversion Short strategy and once again just shortly to the correlation table.

What's important is to compare our new strategy which is pump and dump to trend catcher long, because this is the one that is making the most profits here. And thanks to adding these pump and dump strategy that is only making 17% per year, which is not that much, it's pretty low performance on crypto honestly.

Just by adding this strategy, we pushed the whole portfolio to 220% with a drawdown of 47%, which is again almost the same that just one strategy delivered.

So that's the key of portfolio trading. Once I have more strategies, I can think about my targets, if it is lowering drawdowns or pushing performance, then I can really play with weights as you Richard were rightly asked, then I can really think about the target. Is it performance?

Is it volatility of my trading account? At the end of the day, it is then more up to the trader and what they want to get from their trading. And just to sum it up, how you could push the idea of portfolio trading even further because I'm really talking about portfolios because I think that's the key and I think that not many retail traders are thinking about that much and they should much more than about single strategy,

because portfolio is the key.

How we are for example building portfolios is that we are building sub portfolios first. Let's say we are building some sub portfolios of long trend strategies only, short trend strategies only and then we put it into trend only sub portfolio, which is the blue one that is again built of long trend strategies and short trend strategies.

Just for you to know, long strategies here are from six trend strategies and short strategy sub portfolio or short trend strategy portfolio is built of three strategies in this example.

Of course, we will still have drawdowns, why? Because if the market is not trending, we cannot make money with trend strategies. But that's why we also have mean reversion sub portfolio in the whole portfolio and that's something you should always have in your portfolio, why?

Because especially if you put more mean reversion strategies into the whole portfolio, these strategies are really pushing the stability much, much higher. Of course, the profit is much smaller just compare it here, the mean reversion portfolio is about let's say 500% and the trend portfolio was up to 20,000%, which is huge but it's just crypto on stocks it's the same but just much more results.

And if you put these strategies, these mean reversion long, mean reversion short, trend long, trend short through whole portfolio, you are getting results or performance that is not possible to made by single strategy on any asset.

And if you put it on crypto, the performance is on completely different level because thanks to algorithmic trading, you are taking advantage of portfolio trading, you are taking advantage of a major asset as crypto and you are taking advantage of the third key and its compounding effect. It is something that again traders often forget because they don't think over the long term. It's really about the long term thinking.

Once you think long term and you can compound day by day if the portfolio is growing, then on something like crypto the results can be just tremendous. And yeah, that's it, that's all to the presentation. The only thing for example, to see is the correlation table of the whole portfolio, which is just for your interest to see that the key again to get to stable results is pretty low correlation.

Correlation is really the key and you should concentrate on it as much as possible and once your correlation is slow, you just have to put these strategies together and work on the portfolios. Yeah, and last thing, maybe what we could think about is this return to drawdown ratio. It's not about showing off, it is more about showing that we are definitely in some stage when crypto is right now and it will nothing but get in smaller.

It's pretty sure, it's pretty obvious because if something is giving potentially such a huge result, then the only thing just remember this immature to mature cycle, the only thing that can happen and we can see it, thanks to our institutional clients that it's pulling new capital into the market and the new capital is pushing the liquidity and it's basically pushing this inefficiencies lower. Yeah, so that's it.

Richard Moglen 50:52: Yeah, excellent. Yeah, first of all, this has been really fantastic so far. I think you're able to explain a ton of concepts really, really well.

Definitely want to dive a little bit deeper into a few of them. First things first, looking back, you mentioned this very clearly, the kind of differences you can get from back tested strategies versus what you'll actually get going forward using it. How much do you guys track at Robuxio, the differences between like how your portfolios are performed going in the past versus how they're actually doing in the current day?

What's kind of the difference of performances that you guys typically see?

Pavel Kýček 51:36: That's great idea. Honestly, in crypto, it's slightly different compared to stocks, for example, because in stocks, want to see the performance very similar to the backtest in crypto because you are getting through this immature to mature cycle, the strategies can deliver small results just by this cycle and just by crypto being a little bit more liquid and a little bit less inefficient.

So how we are approaching it is a little bit advanced thing but maybe interesting for other traders is that every single strategy that we trade has some benchmark that we created by ourselves.

It means that if I trade trend long strategy, I typically have another two to four strategies in the portfolio, not in the portfolio but I have them built and I create benchmark of them and I can compare the strategy we are trading for our clients to this benchmark and I can see if it is starting underperforming compared to the benchmark or if it is in line with it or if it is making even better results.

But in general what it is about is understand the strategy because for example many traders do have some drawdown threshold and if the strategy is getting below the drawdown, you just switch it off. Honestly, I think it's nonsense, it could be used if you really don't or if you start with algorithmic trading then you should have such a threshold But if you build your strategy with so called idea first approach,

which means that you know in which phases the strategy should deliver results then for example, you know that if again, we are trading trend long strategy and the market is in some prolonged period, when there is no long trend, it is normal that the strategy can go through deeper drawdowns than in a backtest and it's perfectly normal.

What will be problem if the strategy will start delivering deeper drawdowns in the growing market for example.

So yeah, not sure if I answered your question but it's more about understanding the solution, having the benchmark, benchmark is super important, you should always have some kind of benchmark to compare it with and then it is about also metrics like expectancy and these really hard metrics but these are not the most important ones, more important is to understand what the strategy is doing compared to the market.

Richard Moglen 54:37: No, think that did answer my question, so thank you. So this next question I really like to ask pretty much everybody who I get the chance to interview, discretionary traders or people who I've talked to, but I think it's a really interesting one and I'm definitely interested in hearing your guys approach for this. The question is, how do you define and also manage risk in the markets?

Pavel Kýček 55:02: Yeah, that's great question. That's Yeah, one of yeah, that's one of those questions that we could probably talk about that are the two Yeah. But okay, so firstly, we have to define what's risk in general, what's risk in the asset that you are trading.

For example, in crypto, there are some risks that are not anywhere else on any other asset or at least they are minimized like risk of having some allocation on crypto exchange. That's one of the biggest ones. On the other side, what I'm saying that once all the risks will be covered, the opportunity will be much lower.

So let's define some risk. First is crypto exchange, another one that you are overexposed to one coin that will go to zero overnight without any liquidity. In crypto, it could happen.

Another one is just risk of huge movement during a highly volatile environment and you would be over leveraged, let's say. So once you have all these risks, let's say specified, you can start covering them. Risk of crypto exchange, you can trade on many different solutions or many different exchanges and spread the risk a bit.

Risk of too big allocation, well, trade portfolio of many strategies. And for example, what I didn't mention that every strategy that I even showed you is trading between 10 to 20 positions. So at the end of the day, every single position has allocation between one to three percent maximum, so without leverage.

So again, the risk is reduced quite a lot. And the risk of, let's say, being too overexposed in very sharp movement, let's say to the short side, these minus 30% days they can happen in crypto. Those are for example, in our solution covered again by portfolio approach that way, that we do trade breakout shorts, again, against our trend long strategies.

So we can get to drawdowns on trend long strategies. And we don't need to use stop losses, which is quite a big topic that not well, depends on the asset and leverage that you trade. But the fact is that purely by stats, stop losses are making your performance worse.

Usually if you make such a test that if you would use stop loss and you get out or on the stop loss or you get out at the close of the bar if the stop loss is triggered, you always or almost always get better results without having this fixed stop. Why? Because of intraday mean reversion characteristics of the market, but of course, if you do trade with leverage, you always have to have your stop loss in the market.

But I'm just saying that many traders are just reusing the same tools without proper research. Yeah.

Richard Moglen 58:27: I think that's an excellent answer to the question. So thank you. Thank you.

I think that's great. The next question, it's it's also one I'm sure we could talk about for four hours, but maybe just from a high level, say my my question is basically coming up with a new strategy. So maybe we could pick a trend long, for example.

What would be the first steps when designing a new strategy when coming up with an idea that you guys would take to define the strategy as well as test immediately if it's viable and something that should be pursued further?

Pavel Kýček 59:07: That's another great question. Well, if I should give general recommendations, know, let's keep it like a high level I would recommend going with research first, don't try to find something just by putting indicators in your backtesting platform and trying to find some positive results on the past data, because those strategies will highly probably lose over the long term.

So let's start with some research or with books, some very old ideas from traders from 1980s, 1990s many of them still work pretty nicely.

So do take advantage of it, we do have many strategies in the portfolio that have some roots in these old strategies. Of course, you would need to make some tweaks. The tweaks are usually based on the asset you are trading.

For example, let's reuse the idea of trend catcher. This is basically basic trend following strategy that is based on moving average 50. You just enter if the asset is above close if it is below.

It's working on many assets. Crypto is quicker. So by logic it makes sense to go with a period slightly lower to let's say moving average 20.

And that's it, that's the idea and the idea should be always based on some logical base like trend and mean reversion. Momentum, breakout, mean reversion. These are the three ideas that you should always build your strategies on.

Never try to build strategy based on your discretionary approach. This is probably the worst way how to build strategy is I was there too, that we are trading discretionary and you just put all the indicators or all the price action into some code, this is usually not working, you really should push it to very simple code and then you have to start making many, many, many robustness checks.

For us what it means is that we are trading on the daily data, so we will build our own daily data based on offsets.

It means that we are making our own 24 daily candles basically on every other hour, it's advanced but what you can do is that you can start making these robustness checks on eight hours, twelve hours. Don't go to much lower time frames, because then the logics don't have to work just again, by logic, because the lower you go with your trading approach or trading strategy, the more the market tends to mean revert.

Maybe my biggest and that's the last idea I would recommend is start with research first,

and start with testing models without trying to find viable trading strategy because through basic models, trend ones, mean reversion ones, you can really understand the market and what it is doing and just then you can start building viable trading strategies because without understanding the market, it's pretty high probability that you will fail over the long term.

Richard Moglen 1:02:56: I think the simplicity equals robustness slide that you went through, that's a really key point. That's very important.

Pavel Kýček 1:03:07: Like start with sound idea, momentum breakout, mean reversion, keep as little conditions as possible. Some of our strategies do have just two conditions. For example, the whole strategy is two conditions only and it's performing, it's delivering profits.

So as little conditions as possible and then run robustness checks based on yeah, robustness checks we could go over robustness testing probably over many hours so.

Richard Moglen 1:03:38: No, that's great from a high level and it's super cool to me how even combining two different strategies, a trend with a mean reversion or a trend short with a trend trend long, you can combine their individual metrics and it yields just completely different results that combine this. So that's super cool to me just as a concept. But but first, that might that might be going off track a little bit.

But first, one of the questions I had was, what once you've got an individual strategy, what are some key metrics that you guys look at? And also, there any data visualizations that help you analyze a strategy and see if it has merit, see if it should be explored further and to add to a bucket of trend strategies and then eventually to the overall portfolio construction?

Pavel Kýček 1:04:29: Yeah, great question. Another one, so for me, one of the most important metrics is expectancy or average trade. Why?

Because the higher the expectancy, the higher basically the pillow or the more exactly potential profits you have before your strategy will start in like reducing profits because of slippage, because of fees and so on. So definitely expectancy for some it is sharp ratio.

For us, it is not that important because if you, for example, focus on Sharpe ratio, you would have to immediately exclude trend following strategies, but these are the most robust ones.

So it really depends, because I do understand why Sharpe ratio is like industry standard and many institutional clients want to trade the high Sharpe portfolios because these deliver more stable results and for many it is important but I wouldn't say that for retail trader it is the most important ones. So expectancy and then I would think that it is more about why you build your trading strategy,

because if you build your trading strategy for the purpose of being the part of the portfolio, it's slightly different than if you just build single strategy. Let's say that we have our TrendCatcher Long, which is catching long trends basically and let's think what's the other strategy that would give us profits in very different market conditions.

Right. Biologic trend, something that is making money on short trends, right? So then you start seeking for strategies that are able to deliver profits in these conditions and I don't care that much about single parameters if the expectancy is high enough, that's very important.

Expectancy is very, very, very important for me. And once you have these two in the portfolio, then again, what's another strategy that is giving you the most benefits for the whole portfolio? So once you start building portfolios, the game is changing a bit and you are not going over the single parameters.

Of course, it has to have some basic parameters to be able to deliver profits over the long term, but then it is more about market phases when the strategy is able to help the whole portfolio.

Richard Moglen 1:07:29: Yeah. And just to kind of as a clarifying question and to make sure I'm understanding you, say you have a trend strategy that's acting well. Actually, we've got kind of examples on the screen here that the yellow one here during the period kind of in the middle, it's pretty much going flat Yeah.

Or or declining. You would look to find a strategy that during that same period, those same kind of conditions, market conditions, market regime a strategy that complements that is actually increasing during those periods. And then when the trend strategy is increasing, it's uncorrelated and not pulling back during that same period.

Yeah. Exactly. That's exactly the explanation.

Pavel Kýček 1:08:10: You are basically just trying to put together uncorrelated strategies. Basically the correlation coefficient is something every trader, especially algorithmic one should think about a lot and should understand very deeply and should start thinking about it because thanks to uncorrelation, you can see that trend strategies are losing here, mean reversion strategies are making some money. Why?

Because the whole market was going sideways, it was basically going up, down, up, down. In this environment, it's not possible to make money with trend strategies, but it is possible to make money with mean reversion strategies. It still means that you will have drawdowns, but those drawdowns, the length of the drawdowns are reduced by yeah, many axes compared to single strategy trading.

Richard Moglen 1:09:06: Yeah. Excellent. And this is gonna be a tough question as well.

So sorry in advance, but it's a fun topic to talk about. It's it's kind of easy to wrap your head around when trading to just using two strategies and allocating capital between them when the rule the code that switches one on, the other one is off.

But when you're combining a mean reversion with a trend and all that, from a high level, how are you guys thinking about allocating capital to make sure that the combined performance is as best it can be?

What what do you guys do to think about that? How to allocate how much of your portfolio to one strategy versus another when they've got kind of vastly different rules that turn them on and off? That's another great question.

Pavel Kýček 1:09:56: There are few key things that you have to think about if you put together the weights of single strategy in the portfolio. One is correlation again, it means if you have two strategies into the portfolio, and those two strategies are a bit or even more correlated and you still want to keep them in the portfolio

because of some reason, Naturally, you should push the weight of these two strategies lower because they will tend to deliver drawdowns in the same periods and that's something you want to get rid of. So that's one thing.

Another one is volatility. Generally speaking, the more volatile the strategy, the less the weight in the whole portfolio should be. And the third one is how your portfolio or when your portfolio should make profit or should make profits the most.

For example, here you can see that even though mean reversion strategies are making money, the whole portfolio is still in a drawdown. Why? Because we put slightly higher weight to trend strategies compared to mean reversion strategies, because crypto is trending asset basically.

You want to extract profits from trends mainly and you want to use mean reversion strategies as stabilizers of the whole portfolio. But we have even clients who want to trade mean reversion only or want to trade like only short term breakout strategies and mean reversions to keep the stability higher.

So then I would say that it is also about clients needs or traders needs, let's keep it this way and think about what do I trust more, what are the main characteristics of the asset I trade and what's the volatility I want to get from my trading solution.

So it's hard to give like very very simple answer to it because it's really really more more advanced but yeah volatility correlation and the output of the portfolio is the key things that you should think about if you are playing with weights of the portfolio.

Richard Moglen 1:12:34: No, I think that's a good answer. And I also wanted to ask just because this might be useful for some people watching, Talking about your personal story, how did you first get into algorithmic trading? Did you start discretionary and then kind of make the switch?

Yeah. And what what was that kind of process, all like?

Pavel Kýček 1:12:54: Yeah. As you said, I started as typical retail traders, so it was really through discretionary trading. I was trading on the start like price action formations, I was losing money for some years, then I made the switch to, I would call it semi systematic approach in the sense that I was measuring market movements of stock indexes, NASDAQ one hundred was the most tradable ones.

And I was trading basically mean reversion strategies on three to one minutes time frames if the market was overextended and I was using level two data for it. So that was the first time I started being profitable, it was after quite quite a few years and yeah, it was working next to it, I was working on I would call it systematic investing through these simple rotational strategies on stocks and on ETFs.

And like nine years ago, I started making the switch and I finished the switch after my daughter was born like eight plus year ago because honestly, discretionary trading is super psychologically demanding at least for me,

but as I could see on many of my former trading colleagues, most of them made a switch to algorithmic trading too because yeah, it's hard even especially if you are like pushing the capital, you probably even want to start trading other people money at some level and then the game is on another level from the psychological point of view.

Richard Moglen 1:14:59: Yeah and I'm kind of curious as well, what is a typical day look like for you in terms of what you're doing day to day because the models are doing its thing, how much are you researching, analyzing real time inputs, what was maybe kind of walk us through a typical day and what you're doing to improve the overall portfolio?

Pavel Kýček 1:15:21: Yeah, well, so right now the crypto trading, what we do at Robuxio, it's fully automated. It's not on my stock portfolios. This is semi automated in the terms that I do run my scanners, I do get these entries exits, and just send it to Interactive Brokers one per day.

So the trading part is really covered very, very simply thanks to our trading backend. But honestly, being completely honest, I wouldn't be able to build the trading back end by myself because it took the whole team two years and we are still building it. So I think, sorry it's not completely your question,

but one of the advices or recommendations I would give is if you want to start with algorithmic trading, don't try to automate everything because the automation that should be the last step and it's by far the most demanding ones because you have to understand strategies and then you really have to properly understand the code and that's really not simple.

So yeah, back to my trading day, it's more about research, it's about going through some research studies, going through Twitter too because you get some good ideas there and work on my trading models and just testing, reading and then also what I do right now is having a lot of contacts with institutional investors and also retail clients because they are firstly our clients of course,

so I want to be in touch then but secondly, I try to understand what should be the ideal portfolio for them because if you don't trade such a portfolio that you are able to stick to as a retail trader or as an institutional trader doesn't matter at the end of the day then you simply won't stick to it over the long term and just over the long term, the statistics are playing out and we can make money over the short term, it's just pure luck.

So yeah, it's really it's about testing, reading, testing

Richard Moglen 1:17:52: and talking to clients. That's my whole day these days. Yeah, perfect.

And this this next question is kind of coming back to the training strategy a little bit, so sorry for jumping around. But say you've got a particular strategy or a group of trend strategies that are underperforming the benchmark based on what you guys have set. What are the steps you guys take to try to see what's going on and see what improvements you can make?

Pavel Kýček 1:18:18: Yeah, so what's important to say is that we are trading from the daily data, So if it is underperforming for a month, it doesn't mean anything basically. We are talking about waters here of underperformance. So if I see such an underperformance, I start thinking about why and if it is something that is still within the trading model or not.

If it is not, I can just switch it, I can just start using different trading strategy and if I fit or what I can do is that I will add similar strategy and I put the weight of the worse or the one that is underperforming, I put the weight slightly lower and put the other one next to the old one with again smaller weight and I'm just looking at both of them.

So I think that fixed thresholds are good for the start but once you really start understanding the market and your strategies, it is more about managing them on the month to month basis or quarter to quarter basis and you really have enough time to think about the whole solution, that's another reason why I want to trade from dailies and it is that you have a lot of time to think, to make your research, to make some changes if it is necessary but I'm not making changes very often these days like one strategy a year.

We are adding strategies to the portfolio as the research is continuing, we are always having some strategies in incubation and adding them to the portfolios if they are putting some uncorrelated benefit to the portfolio, that's something we always do because the more strategies the better, but I'm not switching off strategies that often.

Richard Moglen 1:20:40: Yeah, and for thinking of somebody watching this who maybe wants to dive deeper into this based on what they learned today, what resources would you kind of point them towards? What resources did you find helpful personally in learning more about how to do this, construct portfolios, test strategies? What would be maybe some books or other types of resources that you would point them towards?

Pavel Kýček 1:21:05: Yeah, in general, I would start on Twitter, I would start finding or searching for algorithmic traders with some performance that are measurable like Nick Radge would be a good start that I would definitely recommend or Laurens Bensdorp could be a good start too. So these are the guys I would start with and then it is really more about going through research papers.

Just start researching in Google momentum breakout model, mean reversion model research paper, that simply and yeah, honestly that's not the most exciting part of trading but usually algorithmic trading is all about research and the better you make your research, the higher the probability that you will make profits in the future.

You should love it, otherwise you won't stick to it.

Richard Moglen 1:22:18: Everybody just wanted the shortcut there that you're gonna give them and instead you said go read research papers, but that's where putting in the work makes a difference.

Pavel Kýček 1:22:27: Yeah, of course, it's a lot of work for sure. Yeah,

Richard Moglen 1:22:31: Excellent. I really, really enjoyed this. Thank you so much for your time.

Just one last question. If you had to give advice to people watching this, it could be trading wise, could be life wise, systematic trading wise, discretionary wise, what would you kind of, say to everybody as kind of one last parting message?

Pavel Kýček 1:22:51: Okay, avoid one mistake that is connected to trading and it is don't start trading, I mean with your money, real money before you understand the markets, that's the key.

You have to understand what the market is making on lower time frames that it has like mean reversion characteristics, you have to understand what forex market is doing, what commodity markets are doing, what stocks are doing, what crypto is doing and once you really understand the main probabilities through simple models, through research doesn't matter, just then start thinking about trading.

Don't skip this phase because otherwise you just don't know what you are doing in the markets.

Richard Moglen 1:23:47: Yeah, excellent. Yeah, this has been really great. Thank you again for your time for putting this together.

I think everybody watching it, it should have enjoyed it and got a lot of golden nuggets out of it. Even if you're not planning on, setting up algorithmic trading, there's a lot of truths here that apply to designing one one system, one strategy, one discretionary system. So, Pavel, thank you very much for your time.

Please, everybody in the chat, say say thank you to Pavel. And, yeah, thanks so much. If you guys are enjoying this, go ahead and leave a like down below.

Subscribe to the channel.