Generally speaking when we talk about Artificial Intelligence applied to stock trading our mind goes to algorithmic trading and black boxes: but machine learning algorithms and trading algorithms are completely different concepts.

A machine learning model purpose is to recognize price patterns with a given accuracy: it is not a black box, because the model is open and the logics are clear: in our research we use the same logics of weather forecast methods.

A trading system is finalized to provide financial results given specific entry-exit rules: they can be black-boxes when entry/exit rules are for some reason hidden.

Anyway, traders may have several trading techniques that can be automated plus … something else you can call instinct: there is no way to program it, believe me!

Let me give you an example.

Let’s assume we have a machine learning model able to provide a daily prediction for tomorrow price movement on BTCUSD with an accuracy of 80% . (see here)

And let’s assume the model prediction for tomorrow is:

“BTCUSD tomorrow has a 80% to make a pattern Higher High/ Low”

What to do with such information?

Regarding the entry the trader has several options:

• We open a long trade as soon the model provides us with the prediction
• We open a long trade after an intraday pullback
• We use such prediction as a decision support tool in order to trade the same instrument or other correlated ones.

Also for exits there are several options:

• We just exit on next close
• We place a target on the current High
• We place a stop loss on the current Low and a target on the current High
• We manage this long BTCUSD position inside a portfolio
• We use the long prediction to exit some short existing ones

As you see, Machine Learning and Algorithmic trading are 2 different things. You can use both together or you can use them separately.

Why do we necessarily have to automate 100% of the trading process?

According to my personal experience, only brokers benefit from a 100% automation.

Why not to include the AI predictive model inside our own trading decision support tools in a AI-human blended approach?