In our previous article we discussed together the subject of staking, today we will dive into some very interesting things, especially if you have affinities in programming.
You will have to understand, that today we are going to approach together the subject of the creation of bots. Are you ready for it? Let’s get started!
The first thing to do when you want to create your own automatic trading strategy is to choose a programming language. This is just as important as the strategy because, without development skills, it will be very complicated to code anything.
The latter is very interesting because it only allows you, if you want, to implement machine learning elements in your strategy. If it is well done, it can bring a great added value to your strategy.
You will first need to define your strategy clearly. The idea for a strategy always comes from observing how one crypto behaves in one or more particular situations. But many novice traders have an unclear strategy that fluctuates with market events and their fear or appetite for risk, so in reality, they have no strategy.
Programming makes it possible to define a strategy with precision because there is no such thing as “just about” in computer science. Indeed, “sell when the market has risen by 1%”, does not mean “sell the market after +0.9%, a computer is extremely precise.
Translating your strategy into a flow chart, which you will then transcribe into programming code, will force you to define the detailed conditions for setting up and exiting a position.
To define your strategy, depending on your training, you can use complex mathematical concepts or use technical indicators for less complex strategies.
After building your strategy, you will have to backtest it and this is one of the most important phases in the creation of an algorithm as it is possible to make dramatic mistakes.
Indeed, when you start in automatic trading and you code your first bot it is normal to want a successful strategy. It is not uncommon for a beginner to choose to optimize his parameters on all the available data. The result is an algorithm that is overfitted to the data it knows and that performs poorly in real market conditions.
To avoid this, keeping it simple here, you need to separate your data set into two distinct parts and optimize your algorithm on one part. Then, without touching your parameters, you test your algorithm on the other data set. If the algorithm behaves in the same way, then it will work in a real market situation.
In conclusion, algorithm creation is a very exciting thing and you don’t necessarily need a super complex strategy to start.
The most important point will be the respect of the backtesting rules, otherwise, you will be exposed to bad surprises in the market.