Diagnosing Unchanged XGBoost Backtests and Equity Curves
Summary
This short BigQuant forum post concerns a modified XGBoost strategy whose equity curve and recent trade records appear unchanged despite adjustments to the data. The response offers two possible explanations: the backtest module may be reusing cached results, or the model may continue ranking the same stock first when the portfolio holds only one name.
The suggested diagnostic is to change the backtest’s initial capital to force a refresh and, if necessary, make a larger change to the input data to see whether the selected stock changes. The post does not include the strategy code, model configuration, data transformations, or a confirmed root cause. Its suggestions are troubleshooting hypotheses rather than a general account of XGBoost behavior, and neither explanation is verified with a reported test or result.
Key ideas
- An unchanged equity curve after data edits may reflect cached backtest output.
- Changing initial capital is proposed as a way to trigger a new calculation.
- A single-stock portfolio may keep selecting the same top-ranked stock despite some input changes.
- The post gives possible diagnostics but does not establish which cause applied.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.