Training One-Direction Machine Learning Strategies for Gold Trends
Summary
The article adapts a machine-learning trading workflow for assets where long and short opportunities may be unevenly distributed. It labels only trades in a chosen direction, using a randomly selected holding period and a markup threshold to distinguish potential entries from points to avoid. The model predicts whether to trade in that direction, while a second meta-model filters trades; a custom tester evaluates exits from model signals, stop loss, or take profit.
The author describes applying the approach to gold and using a forward test in MetaTrader 5, but the available text gives no detailed performance figures. The method is presented as one possible approach rather than a universal solution. A one-way strategy depends on the trend continuing, so the author stresses monitoring broad market direction and adapting when conditions change. The article also refers to earlier work for background on causal inference, cross-validation, and testing.
Key ideas
- A one-direction labeler marks potential trades for either buying or selling, rather than learning both directions together.
- The label sampler uses a randomized bar horizon and a markup threshold to define candidate trades.
- A custom tester can close positions from model signals, stop loss, or take profit.
- A meta-model filters candidate trades using classification errors from cross-validation.
- Unidirectional strategies can be vulnerable when the prevailing trend changes.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.