Selecting Diverse Strategies for a More Robust Trading Ensemble
Summary
This article examines how to combine technical strategies using weighted votes and genetic optimization. In an earlier experiment, the optimizer selected two strategies whose returns were highly correlated, undermining the intended diversification. The author suggests the optimizer may have favored combinations that made account-level profitability easier to optimize, while acknowledging that the result could have been affected by the small strategy set and a single optimization run.
The proposed revision is to hold strategy voting weights at one while optimizing each strategy’s indicator settings, then assess how to select strategies with less correlated returns. The article describes collecting market and indicator data, labeling outcomes for moving-average and RSI rules over a chosen holding horizon, and using statistical models to compare strategy selection with direct return prediction. It frames equal weights as a benchmark for later work. The discussion is methodological rather than conclusive: it cautions against interpreting one optimization result as proof of optimizer behavior and does not establish that the revised process improves out-of-sample performance.
Key ideas
- An ensemble of strategies may provide little diversification when its members have highly correlated returns.
- The author observed correlated moving-average and RSI strategies selected by a genetic optimizer.
- Fixing voting weights at one can direct optimization toward strategy settings before tuning ensemble weights.
- Strategy comparison uses historical indicators, a holding horizon, and labels based on relative returns.
- A single optimization run over a small strategy set is not enough to establish a general optimizer failure.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.