Today, I have an interesting topic for you, and the topic is artificial intelligence. The article will be a little longer, so I’ll split it into more parts:
- What is artificial intelligence and what we can imagine under this name
- What, on the contrary, is not artificial intelligence
- How does artificial intelligence work in trading?
- Artificial Intelligence in the Czech RepublicČR
A few words to begin with
I decided to write this article for two reasons.
The first is that the name artificial intelligence sounds sexy today, and so it is often used by the sellers of any kind of trading systems. Thanks to this article, you will know what artificial intelligence is and what it is not.
The second reason are the benefits of this technology. At first glance it may seem complicated, but it does not matter. The goal of professionals is to create tools that can be easily used by ordinary people. Therefore, it is a good idea to know something about this topic and to have an idea of what artificial intelligence really is and what is not.
Therefore, the goal of the article is clear and let’s get to the point.
What is artificial intelligence and what to imagine under this name?
A self-adaptable model could be for example called an artificial intelligence. It is a system (I do not say a strategy), which, with every new trade, evaluates the state of the market in the historical context and learns how to trade better.
It is therefore like a discretionary trader who improves and trades with every trade better and better. The good discretionary trader learns from his mistakes and successes instead of doing the same mistake all the time and therefore he is not fulfilling Einstein’s definition of insanity (doing the same thing again and again and expecting different results).
Under Artificial Intelligence, you can simply imagine a computer program that is learning and constantly improving. It also has one great advantage over a discretionary trader – it learns very quickly at the beginning, because it goes through tens of years of historical data, and in a dozen hours it can handle what a discretionary trader would learn in years. It’s kind of a “virtual trader.”
In addition, just as different real traders are, also different models of artificial intelligence (virtual traders) are different. So, there are more ways to work with Artificial Intelligence and you can have for example 20 different “virtual traders”.
What is not artificial intelligence
I consider this part to be very important. As I wrote above, many people consider as “artificial” also ordinary trading systems.
If the strategy is self-optimized based on certain values, it is not an artificial intelligence. The so-called dominant frequency approach is used in trading. We have 3 main frequencies – low, high and dominant. Low frequencies are used for indicators like moving average and are relatively smoothed. High frequencies are typical for oscillators such as RSI.
The dominant frequency, on the contrary, is the frequency that dominates the market and is used, for example, as the value of an indicator. The moving average does not necessarily need have to have frequency 200, but the period can be defined by the dominant frequency, so it is changing with each candle and constantly adapting to the market.
Basically, it’s not a new approach, but it’s not artificial intelligence. However, many people are able to consider this (or similar approaches) as artificial intelligence.
Artificial intelligence in it´s true sense of the word will also not work as a strategy in the trading platform, because virtually no trading platform is designed to work with artificial intelligence. Let’s look at this now.
How we work with Artificial Intelligence in Trading
As I mentioned above, Artificial Intelligence does not work in common trading platforms. And because of their extreme computational demands, these algorithms have to be tightly tailored for the specific hardware they are running on.
For this we have special tools, such as the TensorFlow library, which manages your final code for optimal running on your hardware.
TensorFlow is working with a so-called computational graph that clusters individual operations into higher aggregates and where data flows in a specified order. I cannot program it myself, so I know it theoretically from Pavel Hala. You can see a small part of the computational graph here:
An example of a small part of Pavel Hála’s artificial intelligence architecture
But these tools work separately and if you want to use them for Metatrader or other platforms, you have to do it a bit differently.
Basically, there are three ways:
- You work with signals
The strategy generates signals and you enter to the positions manually. It is especially useful for daily timeframe, so you can execute everything on time.
- Signals are automatically being traded
The strategy generates signals and automatically sends them to the platform. In this case, execution is automatically executed and there is not a problem with time.
- Regular strategy generation
An approach that generates a new file for trading platform with every change in strategy. Usually, a new strategy is created on a monthly basis.
So, if someone offers you a strategy based on artificial intelligence, do not trust him.
Artificial Intelligence on Financial Markets in the Czech Republic
Czech Republic has a relatively interesting reputation in the field of artificial intelligence research for financial markets. I personally know more professionals, but Pavel Hála has reached the peak from my point of view. Paul is an astrophysicist and a very intelligent person. After what I know from him, I can confirm that not everyone should be interested in use of artificial intelligence in trading, but it makes sense to have an idea about this topic.
It’s simple. Although Paul is a very clever person, for whom there is hardly something complicated, he has something to pass on. I like to use the comparison with Nikola Tesla. Nikola Tesla was a genius who invented unreal things. Even if almost none of us understood and do not understand what he was talking about, we use the results of his work (I recommend reading his CV). The task of these people is not to pass on their knowledge in raw form, but to give it to you so that you can use it. Their ideas must be adapted to the tools we can easily use.
The same it is with artificial intelligence, you will meet in places where you would not even expect it. For example, search on Google, client support (sometimes communicating with a robot, not a real person), etc.
Pavel is the SpreadCharts architect, which is a great platform for futures analysis, and it is completely free of charge. Paul and his application made such impact on me, that I have decided joining SpreadCharts as an investor last year, so you can look forward to new interesting features.