Mercor Developer AMA Series — Pentational
Pentational (a.k.a Dorien Herremans) is a developer on the Mercor platform. Besides creating top-level trading strategies, Dorien is a PhD holding professor, teaching data science and artificial intelligence at Singapore University of Technology and Design.
Check out all trading bots of Pentational here:
The Mercor Developer AMA Series
Mercor values transparency and education highly. We like to keep our community up to date and as informed as possible. That’s why we are hosting an official Mercor series: the Mercor Developer AMA series.
Given her extensive expertise and after achieving incredible results, we figured Pentational would be a fitting candidate for the Mercor developer AMA series.
In this article, we will set out last week’s AMA, held in the Mercor community chat. The article will give you an insight into a prime developer on the Mercor platform. The AMA was a great success and gave exclusive insights into the benefits of algo trading. Below is a detailed summary of what was covered in the AMA. Enjoy reading!
It is time for another developer AMA! This time with @pentational! Dorien is a relatively new developer but already managed to have her algorithm enter the top 10!
The algorithm is called Raysor Trade Super Trend and it uses a fully automated trend following strategy inspired by legendary traders like Jesse Livermore, Toby Crabel, and William Eckhardt.
As you may know, the format of the AMA will be as follows: we will mute the chat for the first part of the AMA where we will ask several questions to Dorien. After which we will open up the chat and you will be able to ask questions directly.
The best question will receive a 0.5 BNB investment reward!
Hi @dorienherremans! It is great to have you here!
Hi Mercor! It’s my pleasure to be here today! Thanks for inviting me.
Alright, let’s start with the first question. Could you please give us a brief introduction about yourself, your background and experience?
Well I am originally from Belgium, but I have lived a bit all over the place. I am currently in Singapore, where I am a professor at Singapore University of Technology and Design. So during the day I teach AI, Deep Learning and Data Science, and lead a research group on AI for Finance, as well as AI for multimedia. I have previously worked at the University of London, and was a lecturer in Switzerland as well. But I have currently spent almost 5 years here on the equator. So during the day I work in academia, and during the night I am mostly working on trading algorithms on platforms like Mercor 🙂
Extremely impressive, we have a lot more to learn from you aside from algorithmic trading 😉
I must say, I started my carreer with degrees in economics, but 15 years ago, I wasn’t so keen to join boring industry as a consultant or something, and I decided I wanted to do something much more fun and creating.
Moving from economics to AI, that is not a step you hear often.
I sort of always followed the path of least resistance, and in a weird way it led me here 🙂
In the beginning, I decided to just play around with algorithm for AI and music, and quickly that evolved into multimedia more broadly. In fact, at the moment I am leading the implementation of a Metaverse at our university.
Doesnt seem like a low resistance path to me!
Well it certainly wasn’t always the path of highest salary 🙂
That I can see yes, but you were choosing what you liked best, which is most important.
And actually, I think this is what my passion is. To do something that not many people are doing, to explore new things. Just like nobody was working on music AI 15 years ago before Spotify. The same is true for the crypto market at the moment.
So I find that a few years ago, when I got more into blockchain, I really found a renewed passion for economics.
Alright lets talk a bit more about how you got involved in algorithmic trading, how did you go from algorithms for AI and music to trading algorithms in the crypto industry?
Well the cool thing is that both audio as well as financial data are temporal signals. So you can very easily use the same deep learning algorithms on both of them. So technically the move was ‘easy’. But how did I get into it?
A few years back, I read an amazing book about the life of Jesse Livermore: ‘Reminiscences of a Stock Operator’, I can recommend that to anyone 🙂. It really got me thinking about short term trading and how different it is from investing. E.g. Buffet says ‘buy the dip on the way down’, and Livermore says ‘buy when there is an uptrend’.
Around the same time my wife had a baby. And, being in Asia, that meant that I awake a lot during the night, i.e. during US trading hours.
But soon I realised the real potential is in the crypto market. Which is the modern day equivalent of the unregulated wild west markets of of Jesse livermore in the early 1900s. When I started trading based on some manual signals I coded based on Livermore, and I ended up checking my phone every 2 hours at night. Needless to say it was disruptive : ).
So I worked every free moment to fully automate my trading. It helped that I had a background in advanced optimization problems from during my PhD.
And it sort of went on from there… I made sure I was up to date and followed some amazing courses at NY Institute of Finance, and Uni of Hong Kong. And I try to read anything I can find about those early traders.
What a unique way of looking at financial data and thank you for the reading tip. I am sure many will add that to their reading list.
Really interesting story of how one can self-educate and become extremely skilled in a new field. I can imagine the learning never stops for you.
Looking at your algorithms, what do they focus on and how do you feel your algorithms differ from other algorithms?
Yes you could say that 🙂. I’d encourage everyone to keep reading and learning…
As for the algorithms:
They are mostly trend following. Looking at Raysor Trade Super Trend on MercorL https://app.mercor.finance/developer/algorithmDetail/0xC71F1b67D84A5489FF41B1536500Cd5434443fAa/
This is a special type of trend following. As you are aware, on DeFi, we have to be mindful of higher trading costs. This is a beginner mistake many traders make when backtesting: how much influence could trading costs really have? turns out: a lot!
So, with trading costs not making my Skyital algorithm feasible, I set out to create a slower pace one. Raysor Trade Super Trend uses a fully automated trend following strategy inspired by legendary traders like Jesse Livermore, Toby Crabel (narrow range breakout), and William Eckhardt (turtle trading).
So while there is traditional trend following, we also have a breakout mechanism (e.g. if the market does not move for x days), you can expect a suddent upward movement. All under the right conditions of course (considering hash rates, momentum, RSI etc.
Definitely true, small changes make a big impact over time. A beginner mistake we have seen happening quite a few times.
By trading the ETHBTC pair you have constant crypto exposure while we try to outperform the general market. What you can see for instance, is that when markets crash, BTC goes doen less percentagewise. ETH goes down more. So then we flip to BTC. On the upside, ETH often makes the bigger moves. So when markets go up, we flip to ETH. At the moment, we have a position open 2% increase in the strategy in terms of BTC. The absolute PnL is of course much bigger because BTC itself is going up.
Interesting approach, makes sense.
Could you share some performance history or backtesting results of this algorithm?
Since we are all in the crypto space, we believe BTC’s value will increase right 🙂 So then we just want to increase our BTC holdings 🙂
Let me pull up some backtesting results
Most of us indeed!
Letting Raysor Trade run from 2015 gives us quite the return (remember base currency is BTC).
We can zoom in a bit and start the backtesting from 2018 the bear year.
That is indeed “quite the return”!
BTC’s price (the grey line) went down here. While Raysor Trade kept quite steady.
Very impressive results with very low drawdown.
So even during backtesting, we see longer periods where returns might be the same as holding. But when the market really starts moving is when the profits are made.
I try to make sure the algos are not overfitting. For instance, if we were to see a 90% percent profitable rate. They may mean that I set the oversold/overbought conditions to `overfit’ to those specific traders that occurred. But it wouldn’t work on new data. A rule of thumb is typically 40–50% profitable is great. Too much can be suspicious.
Those wanting to see more details, you can see all the statistics here: https://dorien-herremans.medium.com/raysor-trade-super-trend-2f5b76b1bbe8
Yes because overfitting a strategy would mean, amazing backtesting results, but poor real life results.
Thanks for sharing this!
So this strategy is long only, with daily candles. And you will see that there are very few trades. 24 trades only since 2018. This may be enerving, but you should consider this strategy as a better version of HODL.
That is definitely not a lot, but looking at ETH and BTC, there js probably no need to be trading much more. Especially indeed if you compare it to a just HODL strategy.
How long did it take you to develop this strategy? And are you the only one working on it?
If you don’t count Livermore, Crabel and the likes then yes I am the only one working on it 🙂 It usually takes me a few weeks, during which I will be totally emerged in the process. It will be just me, and books of inspiring traders.
I will test different coings, assets, and markets. Sometimes code that you worked on for a day simply won’t cut it. And you have to throw it away. That’s the nature of building successful algorithms.
Sometimes something looks good for one specific scenario. But then when you widen the timeframe, try a different coin, or trading cost. It just collapses.
So it includes a lot of trial and error?
Yes definitely. I try to make my code modular. So I can quickly add something as a function. If it doesn’t work, I can turn it off again. But in general, the process is very iterative.
Luckily we have great tools like Tradingview to quickly implement things.
Alright, thank you for giving us some more insight in your development approach!
Looking at the Mercor platform, what do you think of the platform, the functionalities, the team and its ambitions?
I love how it democratizes trading. No institutions, not approvals, no minimum capital, no subscription fees.
I feel like there is a constant fight between the established rich wanting to keep their privilege, and us normal people. So the idea of allowing everyone to profit sharing is very powerful. I never could understand if you find an interesting fund you want to invest it, but then it says ‘accredited investors only’. Which has a range of requirements like 200k annual income.
Why should rich people get to have the better investment options?
So while people are trading on centralized exchanges, what Mercor is doing on DeFi is amazing.
That was exactly what we were thinking and what eventually led to our ambition to fully democratize the algo trading industry.
Thank you for the great feedback!
Yes, a great initiative!
I feel fortunate that you included me as your traders!
That brings us at the end of this AMA, if there are any questions from the community, this is the moment to ask them!
M F, [Mercor community member]:
This was a great AMA, thanks! I was wondering, since you are a professor, would it be an idea to work together with Mercor on the Mercor Academy and educate new users on algorithmic trading and AI?
Hi M M, great question! Yes if it fits I’d be happy to be involved in some of the educational content that Mercor offers!
M F, [Mercor community member]:
That would be interesting to see. Thanks again, learned a lot!
To all of you, thanks for having me. If you have any questions afterwards, you can come find me on Twitter @pentational and I would recommend you all to follow Raysor Trade Super Trend 🙂.
Tiberiu, [Mercor community member]
Great AMA 👏
Great AMA, I learned a lot!
Dorien, it was great to have you here! Thank you for the amazing insights, we are all very happy to have you on the platform and are looking forward to your Algo’s performance!
The community got exclusive insights into the methods of our beloved bot creators and got to chat directly with the creators of these profit machines! One of Mercor’s main goals is to make algorithmic trading accessible for everybody by building a bridge between talented developers and expert algo traders and everyday investors. This AMA series is a prime example of doing just so!
The Mercor team enjoyed the AMA to the fullest and loved to see the engagement between investors and developers. On to the next one!