Discussing the article: "Exploring Machine Learning in Unidirectional Trend Trading Using Gold as a Case Study"

[MetaQuotes]

Check out the new article: Exploring Machine Learning in Unidirectional Trend Trading Using Gold as a Case Study.

This article discusses an approach to trading only in the chosen direction (buy or sell). For this purpose, the technique of causal inference and machine learning are used.

Recently, we have been studying implementation of symmetrical trading systems through the lens of binary classification. We assumed that buy and sell transactions can be well separated in the feature space, that is, there is some dividing line (hyperplane) that enables the machine learning algorithm to predict both long and short positions equally well. In reality, this is not always the case, especially for trending trading instruments such as some metals and indices, as well as cryptocurrencies. In situations where an asset has a clearly defined unidirectional trend, trading systems that imply buys and sells may be too risky. And the overall distribution of such trades may be highly asymmetric, leading to misclassification with a large number of errors. In this case, a multidirectional trading system may be ineffective, and it would be better to focus on trading in one direction. This article aims to shed light on features of machine learning to create such unidirectional strategies.

I suggest that approaches of causal inference must be reimagined and adapted to the task of unidirectional trading.

Let's take materials from previous articles as a basis:

• Cross-validation and basics of causal inference in CatBoost models, export to ONNX format
• Causal inference in time series classification problems
• Fast trading strategy tester in Python using Numba

I strongly recommend reading these articles for a more complete understanding of the idea of causal inference and testing.

Fig. 10. Testing only on the forward period from the beginning of 2024

Author: dmitrievsky