What it takes to run automated trading systems

Transcript

Show intro 0:01: Are you looking to become profitable in trading, but you're tired of all those strategies that never stick? Well, you're in the right place. This is the desire to trade podcast with Etienne Crete.

Hear interviews with top traders from around the world who have actually made it. They'll share their tips, techniques, and stories to help you become profitable, stay inspired, and scale up your trading. So welcome to the conversation.

Here's your host, Etienne Crete.

Etienne Crete 0:28: Sitting down today with Pavel Kýček. We spoke, I believe, a year ago when we were in Prague in person — that prior episode is going to be linked in the corner.

You can check it out over here. But, Pavel, good to sit down with you again. Good to talk trading once again, and it's gonna be fun to kinda see where you're at in trading so far.

Yeah. Thank you. Hello, Etienne.

Thank you for inviting me and looking forward to our conversation. Definitely. I mean, tell me what's been going on since we spoke roughly a year ago.

What's what's new and what are you working on now? Yeah. Well, when we are talking together

Pavel Kýček 0:57: last time we were a company of three people working on managed accounts for institutional and high net worth individual clients and also retail traders. We grew quite a lot. Robuxio has 11 full-time employees right now.

We have other part-timers and we are still growing. We are well settled in institutional managed account trading. So yeah, doing quite well.

A lot is going on, but everything seems to be fine.

Etienne Crete 1:34: That's awesome. Tell me a bit more about how did the team expand and the new responsibilities you're trying to fulfill.

Pavel Kýček 1:40: Yes, so we onboarded a new quant. Right now we have a team of six software developers and six software architects. They are taking care of the whole trading engine about the trading backend because not many people realize how difficult it is to properly run managed accounts in a way we are running in crypto.

So it's really very demanding on the software infrastructure And that's why we have the whole team of software architects and developers. We have project manager. We have another quant as I told you. And the team is growing, growing.

Etienne Crete 2:20: So you're at a point where you run these strategies now fully automated. You are running algos for managed accounts, like you said in the crypto market. Any strategies changed in the past year or are the strategies at the core of what you trade the same?

Pavel Kýček 2:34: Yeah, as you said basically all the models we were running when we were talking together are the same as we are running right now. We have added a few new ones, but in general because we are trying to keep the stability of the solution as much as possible, we are not pushing many changes in general. If it is not needed, we try to follow the out of sample out of sample life of our portfolios and models.

That's why we are not making any crazy changes. It's more about how to deal with liquidity because we are well settled in 8 figure territory with our managed accounts. It's about how to spread the liquidity across different coins.

It's really more about like detailed work than trying to find something completely new.

Etienne Crete 3:24: Before we get back to the episode, if you want to jumpstart your day with top stories and tactics, be sure to subscribe to the Desire To Trade Trader Growth email. It's where you'll elevate your mindset, train your skills, and learn to treat trading like a business in less than five minutes a day. It's totally free, and you'll find it at desiretotrade.com.

With the market we've had, I know we had some price and big bull market in the past year in crypto, some more slow market, kinda more sideways. Was it a good thing for you, or do you have any times where you kinda perform a little bit less well than usual?

Pavel Kýček 3:56: In general, it was a good market for us. Why? Because there was pretty big dump at the start of 2025.

And because we are also trading through short side, we have made some nice money on this one and we also grew by AUMs. Thanks to this market because as we are talking about the edge to the short side is much weaker compared to the long side on crypto. Similar to stocks for example, not many professional trading teams are trading the short side too.

And that's why it was our advantage that we were able to lock in some profits during this period, and that's why we grew quite a lot in this period too.

Etienne Crete 4:39: Interesting. Are investors more careful when the crypto is going in a down market? That you feel like they still wanna invest or they're kinda more afraid to go in even though you can still make money in down markets anyway?

Pavel Kýček 4:50: Yeah, that's a very good question. In general, I would say that it depends on the type of investor. Institutions, they don't care that much because they know what the parameters of our portfolios are.

And as long as we are staying in some threshold or under some threshold based on volatility, max drawdown, expectancy, Sharpe ratio and these kind of simple metrics, we are basically okay and they don't care that much. On the other hand, you can see here the difference between individual traders or investors and institutional investors because really individual traders, they tend to panic much more. Yeah.

But it's about explanation, know, the more we are explaining what it is about, the more they understand and the less they are making like some quick steps which are not that great for them.

Etienne Crete 5:48: Yeah. That totally makes sense. I mean, I think we covered this in the last episode we did together, but for someone who's new, I think the toughest thing is gonna be to find that very first investor, the first person that gives you money to trade with or the first one that you manage and account for.

How did you find that first investor? How do you tell people to go about finding their first investor?

Pavel Kýček 6:09: For us it was about socials. We started on X, we started on LinkedIn, we were sharing a lot of information, we were talking about what we are building, what are the basic rules for our models, for our portfolios, how we are approaching trading in general. And that really helped us growing pretty fast.

Etienne Crete 6:31: Interesting. And was there any questions about, like the first person, did they question more or did they are they willing to give you capital to look into how you trade or Yeah, course. The longer we are on the market,

Pavel Kýček 6:44: the higher the stability of our clients. Let's put it this way. And it's also about our growth.

As we are growing, we are not onboarding the smallest retail clients. And honestly, you can very well see the emotional difference between the smallest retail clients and even more advanced high net worth individuals because the main difference is patience. The problem with small retail trader is that they tend not to be patient and it is really the worst, worst approach to trading.

If the patience is not there, you cannot go through drawdown and build your equity curve to all new all time highs.

Etienne Crete 7:33: I think it takes a lot of effort also to just like not touch something when you know it's working decently well. You always want to make things better, you want to try to improve things, Especially with algos, you could go and like go crazy, you could go with like AI and try to like build some more fancy ways to do it. How do you make that balance between keeping things the same versus trying to keep improving also over time?

Because you always want to, I guess, get better over time.

Pavel Kýček 7:56: That's a good one. Well, in general, we know what we are trading and why. You know, we want to be trading models which are traded on stocks, which can be traded on commodities.

We are taking these models and we are trading them on crypto. And now because we are running momentum and mean reversion approaches, we cannot find infinite number of different models. So we already know that finding 100, 200 different models probably doesn't make a lot of sense for us.

That's why we are benchmarking all our live models. We are checking what they are doing, comparing them to historical results, comparing them to backtests and basically based on these numbers we can see that we are okay. And it's not about trying to optimise or re-optimise some model which is running live for a year or two years because honestly the longer the real out of sample period of a model, the higher value the model has for traders.

So that's why we know that optimization is not the way we would want to go.

Etienne Crete 9:07: That makes sense. How much do you consider yourself to be a trader these days? I'm guessing in the beginning you do a lot more trading than you do now.

Do you still go and trade manually or do you let the agos

Pavel Kýček 9:16: run and do their own work? Yeah. So crypto, I'm not trading manually at all.

Of course, Robuxio is fully automated solution, which is completely running in the cloud in Amazon. Me personally, I'm still running some models on stocks, let's say semi automated, which means that I'm running my scanners once a day and sending orders to opening or closing auctions. But with Robuxio, we are also going to

Etienne Crete 9:45: stock trading and commodity trading. So at the end of this year, I would like to have everything fully automated under Robuxio. Right.

Some people might also ask you this is like, why don't you automate the whole thing and all strategies? Do like the aspect of trading manually or do you plan to automate everything in the future? Yeah.

Everything will be automated. For algo trading,

Pavel Kýček 10:07: I would say it depends on which level you are. You know, I'm not saying that I'm super super big expert, but I've been in trading over eighteen years actually. And I've seen a lot of things in the market.

And now I think that for an algo trader, which is starting, it makes a lot of sense not to run everything full automatically. First, because it still pays off to watch the market on a daily basis, what it does, how it reacts to your orders, how it reacts to your open positions, how we are closing the positions and so on. And second, the final step of algo trading, the automation to make it really robust.

Yeah, it's hard. It's not that simple.

Etienne Crete 10:53: Before we get back to the episode, if you prefer to watch content, then go find me on YouTube. I have this exact same episode on YouTube. I'm Desire To Trade / @etiennecrete on YouTube.

Just subscribe to the channel, turn on the notification bell because then you'll get notified in real time. It'll tell you whenever I post new episode, so you never miss anything. Now let's get back to the episode.

Right. Is it possible to make a algo strategy as good as you would trade manually or even better? And how do you try to do that?

Pavel Kýček 11:22: It's a good question. I would say for most traders, it's much better to trade algorithmically than discretionary. That's my point of view.

Why? Because the emotions are a little bit smaller. You tend to understand the drawdowns a little bit better.

And my belief is that trading, long term trading or long term profitable performance is about going through drawdowns. Right? You simply have to survive your drawdowns.

And if you are trading discretionary, you have some space not to understand enough what's going on with your trading. While if you are running some algos, you are trading full algorithmically, you always can go to your tests, you can always go back to our historical data, and you can always reevaluate that the stage you are going through right now is based on the statistics, is based on the expectations.

So that's why I would say that for most, algo trading is a little bit simpler.

But on the other hand, I would say that the best discretionary traders will make higher returns compared to the best algorithmic traders. But we are talking about top out of tops.

Etienne Crete 12:49: Yeah. Definitely. Not not beginners for sure, which is Not a good point.

Definitely not. At the same time, the mindset aspect is still there somehow. Did you ever feel, oh, I should just turn off the algo.

It's like like turning off, it's not working well. And then you need to turn off or don't turn it off. Do you ever feel like you want to turn off an algo completely?

Well, there is always psychological

Pavel Kýček 13:08: aspect in any trading, right? Because one thing is looking at your five years equity curve and go over the drawdown in a minute. And another thing is to go in through three, six months drawdown day by day.

So it's big difference. It's not about that I would be pushed to switch off something or to make some changes. But of course, you are definitely less or it's always not emotionally nice to go over drawdowns, especially if you are managing other accounts and especially if you are managing 8 figures AUM.

It's different level, know. When I was trading for myself only a few years ago, I almost didn't mind going through drawdowns. Then I started trading with Robuxio and that's just another level.

You know, you get used to it, then the AUM grows. You are getting used to it again. So it's a process.

But the more you are in trading, the more you understand what are the metrics you should be looking at, what is the most important part of trading and it's really consistency.

Etienne Crete 14:22: Tell me about the investor relations part of managing accounts for people. Do have people calling you when they have a draw on and they feel bad and they wanna take their money out? Or I'm guessing it's different within with the institutional investors, like big firms who invest in you, but how is that relationship with investors that you have to maintain over time?

Well,

Pavel Kýček 14:40: with institutions, it's really different, as you said with individuals. Honestly, our investors who are trading with Robuxio, these guys, they usually tend to be patient and tend to understand because we are really trying to educate a lot and then they understand what it is about.

And really I have to stress that most of our individual investors are on the high net worth individual side and they already are invested in stocks, are invested in some other investment vehicles.

They understand that the key is long term patience. So they know what's the maximum drawdown, what's the volatility of the solution And then, yeah, we are getting emails from time to time, but it's not definitely that I would be on the call from morning till the evening and explaining anything. That that's that wouldn't be an environment I would be, I would want to work in actually.

Etienne Crete 15:41: And probably not. But how do you handle these worried investors in general who are wondering is he going to lose money forever? Is he gonna come back on track?

What do you tell them to think about? Or do you just let them take out their money? Well, honestly,

Pavel Kýček 15:55: well, it's still their account so they can do whatever they want, right? We are only using our software on their accounts. That's one.

Second, we don't have that many investors who are redeeming. Most of the investors they understand what it is about. And they really tend not to panic.

And that's probably because we are really educating a lot. We are sending lot of newsletters, we recorded a lot of videos in the past. And we are in touch with them.

So they do understand what's the solution about. And if and I think now we are getting back for example to algo trading versus discretionary trading. If the understanding is good enough, then people tend to stick to the given solution.

And that's why they understand what our algorithmic solution is about, they tend to stick to it much more. Honestly, we have had quite a few clients who were trading discretionary big numbers, really big numbers up to 7 figure accounts. And right now they are more like, yeah, it's pretty cold, like the account is almost not moving during the day because they were used to crypto volatility of fifteen, ten, 20% per day to the upside of course,

but also to the downside and with us all the portfolios are made that way that they are more, still more conservative than aggressive.

Etienne Crete 17:27: That's a good point. And sometimes the worst part is if they don't know anything, like they don't know about what's happening in the market. And they see they can't go down but they don't know why.

If you explain them and you kind of educate them, like you said, I think it's a good thing to do for sure. Yeah, for sure. I think in general, I would say that many traders, retail traders,

Pavel Kýček 17:45: they end up trading because they don't understand what they are doing, why they are trading, what they are trading, and they are not having the proper understanding of fundamentals.

Etienne Crete 17:56: Tell me more about fundamentals in terms of algo trading.

Pavel Kýček 18:00: Yeah. So it's more about the fundamentals of the market in general. So you have to understand the market first, right?

For example, crypto market. If you should describe the crypto market a bit, we are talking about momentum long market which has mean reversion long characteristics. It's oversimplified.

It means that crypto market has much bigger edge to the long side. It's similar compared to stocks for example, and short downside movements tend to be mean reverted to the long side. It's again similar to stocks.

And you have to have this basic understanding or it's always nice to have this basic understanding first. If you know what's going on, then you can start building proper models on this market. So for example, if I know that crypto is momentum market to the long side, I somehow also know that the best models to start with would be momentum long or trend following long or breakouts to the long side.

And that's by the first step. Then the second step is that in algo trading, you can take advantage of diversification because it is much simple, simpler to trade on many, many coins and quite a few approaches compared to discretionary trading.

And again, now we know that we are trading for example, momentum long or some breakout to the long side, we want to diversify so we can for example split the capital into 10 pieces and run some well diversified momentum long approach.

And that way we can add more and more models to the portfolio, The lower the correlation between these models, the better and we can build something which is much more stable compared to single model trading, for example.

Etienne Crete 20:01: Before we get back to the episode, if you wanna get the behind the scenes of my life as a trader and just how I make it all work while traveling, go follow me on Instagram. It's Etienne C-R-E-T-E. It's all in one word.

It's where I share trading lessons every day, and you get to see what I'm up to in real time, plus what guest I get to bring on a podcast. I would love to see you there. Send me that you would also put less risk on the short strategies for crypto or you could just do best the same?

Pavel Kýček 20:32: Yeah. That's a good one. Well, to the short side, we would have to divide between momentum short or some breakout or trend following to the short side and mean reversion short.

Momentum short is the weakest edge in crypto in a sense that especially the bigger coins, they tend to mean revert and basically trend to the long side much better. Because the shorts really, you are getting whipsawed pretty often, and it can really hurt. So one is to diversify, which means to spread the allocation across 10 — better 20 — positions.

That this way you are reducing the left tail risk of a single position on some small coin. And second is to diversify across approaches also to the short side. And then we have Mean Reversion short, and it's slightly different approach.

It has quite some edge in crypto, but the left tail risk, the risk that the coin will make 300% overnight is still in the market to the short side So what's the simplest way diversification again. So diversification is really the key in trading, would say in general, but especially in crypto and especially if you want to trade something else than just Bitcoin or maybe Ethereum, also coins which can really make 600% during the day.

And it's possible and it's happening.

And of course, if this 600% is going with you, it's perfect and worth is great. But you always have to think about the risks first. So you also have to be always well protected and you have to think about how to protect against this type of risk.

Etienne Crete 22:25: Definitely. I mean, that's good advice for sure. Last time we talked a bit about this too, the fact that people expect you can just code a strategy into an algo and then your work is done and you're kinda good to be retired, like you're good to go and you don't have any work to do.

Tell me about the kind of work you do these days now. Maybe it's different than, of course, not coding all the time, but what is the work involved in running the algo now that it's already put into place?

Pavel Kýček 22:47: Yeah. Well, so one thing is to build a model. Right?

Building the model, especially on crypto, I'm always recommending going with the models which are trading somewhere else Because on crypto, we are still playing with a few years of data. Crypto futures are traded from 2020 — or rather 2021. That was the year when we had some more pairs already and you are getting some more data.

So how many years are we playing with? Right? A few.

So that's why it's always nice to have a model which is traded somewhere else. Then of course, we have to make some adjustments on crypto, but that's just the first part. Building the model is one part And second part is how to trade algorithmically.

And if you want to trade automatically and on crypto, it's much harder to trade semi automatically. For example, some trading on stocks because crypto is 20 fourseven. So automation is almost a need.

And if you want to properly automate the approach, it is really a task. Me personally, I wouldn't be able to do it. So you have to be well protected against API changes, you have to be well protected against some changes in the cloud for example.

So people all sometimes think about algo trading or fully automated algorithmic trading as about something which they just switch on and let it run forever, basically some holy grail of trading. And it's not that far, but it's not also that simple. You have to put a lot of work in.

For example, at Robuxio, just to give you some numbers, we are definitely putting the biggest resources to the software engineering team. Because we know that the edges we are trading are pretty big and they can make a lot of money. But what's important is that the portfolio we have built in the past is the portfolio we will be really able to trade,

which means that really you have to be very sure that the trading engine is doing exactly what it is supposed to do even with any possible changes on the subaccount part for example.

So that's why we are for example having one bot which is running on each sub account, client sub account and it's checking that all the orders, all the running trades, everything what's going on is exactly as it should be based on the benchmark we are keeping for each client on our site in the Cloud Amazon. So yeah, of course our solution is overkill for an individual retail trader.

But in general, people tend to underestimate the automation, I would say.

Or they tend to underestimate more than the automation itself, the risks which are connected with the full automation. You always have to have many checkers in your trading bot or trading software that if something is happening you have to know. And that's a link many retail traders are missing.

They for example are able to do some automation in Python, but they don't spend enough time having all the checkers or the alert system which should be running above the automation, so that you always know what's happening.

Etienne Crete 26:37: That's a good point. Yeah. Definitely.

What so far has gone wrong for you in terms of the algos? Was there any problems or issues that you've dealt in the past that you had to fix? Yeah, that's a good one.

I have to say that fortunately

Pavel Kýček 26:52: we haven't had any major, major issue with the algo itself. Why? Because we have pretty strict process how we are for example pushing any change.

Anytime we are making some not even major change but any change in the software We are always running it on a few of our accounts for a few days. Before running it on our accounts for a few days we are running the solution with for example one hour frequency to get as much feedback as possible And just after this evaluation process,

which takes some time, depends on how big the change is, we are starting trading it on the real accounts of our clients, not before. So because of this and because of one of our co owners, Deneb is great, great software architect, we are always well protected.

But really that's also the reason, the main reason why we are able to do it this way is that there is a big group who is working on this one full time.

But I would say for a retail trader, being a retail trader, I would for example have some very small account which I would be using as a test account and I would push my updates on with any trading engine solution on this small account first to see that everything is exactly as it's supposed to be And just then after some evaluation process, you could start pushing it on your main account.

Etienne Crete 28:36: That's good. That's for sure. Tell people watching this where they can connect with you or find out more about your work after they're done watching.

Sure.

Pavel Kýček 28:44: So you can go to robuxio.com where you can find everything, or to my Twitter account x.com/pkycek. Or what could be potentially interesting for your audience is that I've also made a course based on the course I've led at university. I've had one semester of algorithmic trading at university and I recorded it in English, and it can be found on robuxio.com/course.

Etienne Crete 29:18: Awesome. Put the link below. People can can check it out and reach out to you there.

I appreciate you coming here, Pavel, the advice you have to the listeners. Pretty good stuff. Hopefully, we can catch up with you in the future and talk to you once again.

Sure. Thank you for having me.

Show intro 29:30: Thanks for listening to the Desire to Trade podcast. We hope you enjoyed this episode. Make sure to leave a review on your favorite podcast platform, and don't forget to subscribe and follow on YouTube at youtube.com/desiretotrade.

Thanks again for listening.