Preserving Sell Signals When Filtering Buy Candidates
Summary
The document discusses a backtesting problem in a traditional trading strategy: filtering the data to include only instruments with a positive value for a buy signal improves reported performance, but also changes the sample available to the sell condition. The question is whether the improved result can be reproduced without filtering the data.
The suggested approach is to keep the full dataset, label rows that meet the condition with a binary value, and use that label to rank or select buy candidates. This leaves the sell condition's sample intact. The answer cautions that this changes strategy behavior, so the resulting equity curve will not necessarily match the filtered version. It offers no backtest results or detailed implementation steps; it recommends comparing the approaches experimentally to see how similar their outcomes are.
Key ideas
- Filtering the dataset for buy candidates can also remove observations needed by the sell condition.
- A binary label can identify qualifying buy candidates while keeping the full dataset available.
- Ranking candidates by that label changes strategy behavior and may produce a different backtest curve.
- Compare both approaches empirically rather than assuming they will have similar results.
Tags
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.