Trading Algorithms Outperform Human Traders – Interview with UpBots CEO Benjamin Duval
Whether or not the markets are bullish or bearish, crypto traders always seek opportunities in order to turn their knowledge into real money. With UpBots, Benjamin Duval creates a marketplace for this knowledge. In our exclusive interview, he talks about algorithmic trading and how trading bots can outperform manual trades by humans.
1) Please tell us more about yourself. What was your intention behind creating Upbots?
Sure thing! Well, UpBots is the fruit of many events that happened to me. In fact, trading assets all started when I was 20 years old, so it’s been over 10 years. My father had just passed away, and I found out one day when I went to the bank to open an account, that he had already opened one for me, with shares in it.
3 or so years later I became aware of crypto and found it fascinating, but it took me a couple more years before I began to actively trade it.
And then, I created Crypto-Addicts originally with just a Twitter account. At the time my only objective was to share some advice on crypto and my trades. And something crazy happened, people started following me and asking me if I had a private channel with a subscription… well say no more, I created one.
During that first year of having an offering, we recruited over 10 new team members. I asked Julien Quertain to join me and we made over 1M revenue without any initial cap. And I’m really happy about my collaboration with Julien. While I’m the dreamer, the creator, he is the manager, controlling everything and checking that we are doing things right.
Fast forward to 2019 and Crypto-Addicts was already much bigger than simply me and a Twitter account. However, we decided to merge the company with CryptoMedics to create 4C Trading.
It can be surprising, but my motivation for this merger was to better enable us to offer the best product in the market place and to be a cornerstone in the industry, known for reliability. I wanted to make sure we were an organisation known for transparency because so many corners of crypto are shady. I have always believed that being good is being transparent, and this collaboration with CryptoMedics ended up as very fruitful. We were very strong technically and they had very good traders… together we are now far stronger than before.
I’ve always been driven by a desire to innovate and disrupt so the crypto space is one that’s really interesting to me. Blockchain technologies, Decentralised Finance, Distributed Ledger tech is the most exciting industry anyone can work in right now.
And then … we arrive at UpBots. The genesis of UpBots occurred at the cross roads for a number of needs I saw around me. For one, my traders and our community were looking for a reliable, trustworthy, all in one platform where there was as little friction as possible to the process of trading.
Anyone who trades right now will understand this. If you want to take money out of one asset class and put it in another…well, you are going to need a lot of patience and many browser tabs open. It’s not a fluid experience.
Another need I saw in the crypto space came from me being a crypto evangelist. I want everyone to use it and having forex and commodities and DeFi solutions available through Upbots is a good way to help make it easier for new money to flow into the crypto sector. We need a place where forex traders come, see the possibility to trade crypto as well, and start.
2) What would you consider the most essential skills you need in order to become proficient at trading and creating trading algorithms?
I see there being 5 main essential skills anyone should have in order to achieve this:
It may seem obvious to some, but it is necessary to remember that to develop a trading algorithm, it is absolutely necessary to master basic and advanced concepts such as support and resistance, supply and demand, liquidity.
Numerical and programming skills.
We can be the best manual traders in the world, if we do not master a programming language, it is obviously impossible to create a trading algorithm. Quant also needs a solid background in statistics, signal theory, and big data analysis. Knowledge of machine learning programming is a big plus but is not necessary to create an algorithm.
Think outside the box.
One of the most important skills outside of programming and trading knowledge is undoubtedly creativity. Indeed, knowing how to program a model already well known to the market is certainly a good start, but to really succeed in extracting alpha from the market, it is mandatory to think differently and try something new.
This necessary trait is like the previous one in the sense of calculated risk-taking, trying out a new strategy, for example, based on something new. New is good when it comes to trading strategy.
Again, obvious but necessary for the survival of an algorithm developer. It’s all well and good to program a trading robot, if it can make your entire capital fly away on a mistake, it’s certainly not worth it.
3) Is there a computer language you prefer for creating trading bots? And if so, why?
Well actually, there isn’t really a preferred language because it all depends on what we want to do with our trading algorithm.
Indeed, some algorithms are sensitive to execution speed, such as high frequency trading algorithms, while others are designed for slower trading or for analysing huge amounts of data through machine learning.
For algorithms that are extremely sensitive to execution speed, the C language is the best because it is the fastest in processing huge amounts of information. It is nevertheless one of the most difficult programming languages to learn.
Research and machine learning.
When it comes to research, prototype creation and machine learning, the python programming language is the right one to use. Indeed, it is one of the most complete for this task thanks to its high-performing libraries.
Before putting an algorithm on the market, or allocating capital on it, it is necessary to test it on a large amount of data, following statistically robust testing processes. R is the best language for this, but MATLAB can also be added.
4) Do you believe that trading algorithms have a limited lifespan? And if so, how long does it take on average for a profitable algorithm to no longer become profitable?
Well, as is the case with any mechanical trading system, i.e. based on immutable rules and parameters (at least for a certain period of time), an algorithm necessarily has a lifespan.
Indeed, a trading system, if it is relatively well known and used, will lose its edge over time and will, at the very least, need to be improved or get its parameters changed.
However, even some trading systems based on a simple moving average can be consistent over the long term for the simple reason that their edge is the absence of emotion. This is especially true in an immature market such as the cryptos market, as traditional markets are much more efficient, it is much more difficult to program a winning algorithm on them.
Simply put, an algorithm based on fixed parameters will have a shortened lifespan because the market is adapting and slowly erases the advantage of the algorithm.
5) If trading algorithms usually outperform human traders, why aren’t more people traders attempting to create their own?
The answer is actually quite simple, there are very high barriers to entry. Knowledge of a programming language is the first and most obvious one.
Indeed, to start trading manually, not necessarily profitably, we need proof of residency, a copy of an ID card and that’s it. In addition, there are a staggering number of trading gurus on the internet whose sole purpose is to sell their training, which is supposed to turn you into a Day-Trader in 10 hours.
The reality is quite different, it takes years of work to manage not to lose money consistently in the market, even more so to be consistently profitable.
As we mentioned before, every trading system has a lifespan before it needs to be improved, which makes it unlikely that a winning strategy can be disseminated to as many people as possible for a few hundred euros.
It is way easier to get in the market manually as you don’t have to learn how to code beforehand.
6) How many attempts do you need to create a profitable trading algorithm? Is the process of developing a successful algo a quick affair, or is this something that usually takes months or years?
Well actually, there are several types of algorithms, using different strategies. As discussed above, a strategy that works in one market, such as cryptos, will prove useless in the FOREX market.
The reason is simple, the forex market is extremely liquid and mature, which makes market inefficiencies rarer and harder to exploit as the stakeholders (bank, investment funds, …) weigh heavily.
The first step in the creation of an algorithm is an idea, a way to exploit a market depending on whether it is mature or not, liquid or not, regulated, etc.. This first step in itself can take 5 minutes in the case of an algorithm created by a retail trader but it can take months or even more for a professional high frequency trading algorithm.
The programming, testing and optimization step then comes with a long period of backtesting on the longest possible amount of historical data. Again, this can be done very quickly in the case of a simple algo and can last several months or even longer for the most complicated algorithms.
When this period is over, forward testing can start with a test of the algorithm under real market conditions and should be left running for a sufficient period of at least a few months.
Only after all this process the algorithm can be used in live trading. This process takes more or less time depending on whether you are a retail trader looking for a simple algorithm for personal use (a few months) or a financial institution (months or even years).
7) What if the market is excessively algo traded? Do human traders still have a chance of being profitable? Furthermore, how can you tell that a market is saturated with algo traders?
The market, and more particularly the traditional market, is mainly traded (more than 70% of positions) by algorithms. The main reason is the size of the major players in the classical market such as equities or forex.
Indeed, high frequency trading algorithms allow institutions such as pension funds, which have huge liquidity, to split their trade into several small trades in order to limit the impact of the latter on the price of an asset.
Indeed, by selling their position directly at the market price, without splitting up by the algorithm, the price of the asset could fall and no longer constitute an interesting opportunity for the pension fund.
There is not yet any player of this importance on the crypto market, but we can nevertheless note the presence of market making algorithms to ensure the liquidity of a corner.
Can the human trader cope with such an adversary? Yes and no!
Indeed, the average retail trader will continue to be robbed on the market, it has always been like that and it will always remain so. The stock market is a zero-sum game, you need a loser for every winner.
On the other hand, some traders who are price-wise and who know the principles of market liquidity, after years of studying and testing them, may find it easier to predict market movements if the market is controlled mostly by algorithms.
8) Is algo trading in the crypto market similar to algo trading in more traditional markets such as Forex? What are some notable differences?
We’ve already touched on the answer to this question, in part, when we were looking at the steps involved in creating an algorithm, it’s not the same thing at all.
The main difference is the liquidity available on the FOREX market compared to the cryptos market. A very simple algorithm will be able to work on cryptos because this market is still immature and inefficient, but it will not be able to work with any size of position. It is out of the question to imagine positions taken with a billion dollars on the Binance Coin for example, while it is very common on the EURUSD.
To sum up, it is possible to use simpler algorithms on cryptos while these will not work at all on the forex. Additionally, the rationale behind an institutional algorithm could take a more complex form as there are many more different kinds of instruments in the legacy markets, than in crypto for now.
9) Do you consider that certain trading signals carry more weight than others? Which signals do you focus on when creating a trading algo for the crypto markets, if any?
When creating a trading algorithm, in addition to the basic strategy idea, it is extremely important to choose the right trading pair, asset, and in our case, a coin.
Why is that?
It’s very simple, we prefer to allocate development time to a wedge trading algorithm that is not only buyable, but also shortable (possibility of betting on the downside).
If the algorithm is going to be used by a large number of people, such as our community, we also need to keep an eye on the trading volume of our coin. This must indeed be sufficient to minimize the risk of slippage.
Finally, some pairs have lower fees, such as BNB or LTC on Binance futures.
If a coin has less fees, a sufficient volume and is shortable, we will allocate more time to the creation of an algorithm on it than on another one.
10) Finally, how will UpBots make algorithmic trading available to the retail investor?
One of our goals with Upbots was to create a trading environment that levels the playing field for newbie and retail traders. We have a lot of features to automate trading and to de-risk trading if you are starting out, staring at the mountain of learning that is technical and fundamental analysis.
On Upbots anyone, even those with no coding experience, will be able to set up their own trading bot.
That bot can copy trade a pro traders algorithm from the community (anyone can create and effectively rent out their algorithm). Also they can follow a pre-set trading strategy (for example swing trading BTCUSDT), or they can create a bot that integrates with their favorite signals provider.
And for those with coding experience, they will be able to create it on Tradingview and link it to Upbots, or in the very near future, they will be able to code it directly on upbots in the “Algo Lab”.
Regarding the renting of bots or social copy trading, for example, the best part is this is free as part of the platform. Fee’s get paid only when the trade is successful.
But there is more if you want to rent your algos, or if you want to be a “master trader” followed by others. Because the platform is on the blockchain, one of the fundamental features of Upbots, is that successful trades will be rewarded if and when others are following them.
So actually it’s an incentive for pro traders, quants and signal providers to create and rent out their algorithms, and for those algorithms to perform.
We’ve worked very hard on the tokenomics for the project with some of the smartest minds in the business to create a trading eco-system where this value exchange works really well and people get rewarded in a fair and equitable way.
So for the retail investor who wants to dip their toe into the turbulent trading waters of the crypto markets, Upbots will offer a safer, next generation option that will support them no matter what their skill level is.
If any of your readers want to find out more about that, or have any questions all of the awesome Upbots team, including myself are available in our Telegram group here: https://t.me/Upbots