Signal Filtering Methods and Their Backtest Trade-offs

Summary

This article examines filters as rules that suppress some of an automated strategy’s signals, dividing them into bandpass filters that select conditions over a range and discrete filters that accept signals according to a pattern. It demonstrates the process on a EUR/USD minute-chart Expert Advisor, testing filters based on candle direction across several timeframes and combining those with a discrete condition. The presented reports compare profitability, trade counts, average outcomes, and drawdowns before and after filtering.

The examples suggest that filters can change a system’s trade frequency and performance profile, with different choices favoring profit measures or drawdown. The author emphasizes that a filter suitable for one strategy may harm another, and that historical improvements do not guarantee future effectiveness. The sample is limited to a particular system and historical period; its reported modeling quality is also low, and the article explicitly warns against using the example system in live trading. Filtering should therefore be treated as a hypothesis to test, not a reliable shortcut to profitability.

Key ideas

• A signal filter removes trades according to a condition, and the article groups filters into bandpass and discrete types.
• The examples test candle-direction conditions across timeframes and combine a time-based filter with another rule.
• Filtering changes both the number of trades and performance measures, so evaluation should include more than net profit.
• A filter’s effect depends on the strategy and historical sample, and past improvements may not persist.
• The example backtest has limited evidential strength and is not presented as a live-trading system.

Tags

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.