Every filter you add costs you monthly return
Adding more filters to an algorithmic strategy does not necessarily improve its performance; in many cases, it actively degrades it.
Adding more filters to an algorithmic strategy does not necessarily improve its performance; in many cases, it actively degrades it. My recent tests on a breakout strategy for four currency pairs confirm that only one specific filter, the Higher Time Frame (HTF) trend check, provides a genuine quality boost. I tested various combinations of filters (HTF, ADX, RSI, SMA slope, Efficiency Ratio, and Level) against an out-of-sample (OOS) baseline PF of 1.39. PF (profit factor) is the ratio of gross profit to gross loss; a value above 1.0 indicates a profitable system.
| Configuration | OOS PF | Win Rate |
|---|---|---|
| Baseline (Plain) | 1.39 | 37% |
| HTF (Daily Trend) | 1.48 | 38% |
| HTF + Slope | 1.49 | 38% |
| ADX (20 threshold) | 1.27 | 37% |
| All 6 Filters Combined | 1.38 | 37% |
| The data reveals three clear takeaways. First, the HTF filter, which ensures the hourly trade aligns with the daily trend, is the only reliable addition. Second, ADX filters were counterproductive, consistently lowering performance across all combinations. Finally, stacking more than two filters led to diminishing returns or outright degradation. The idea that “more filters equals better results” is a myth. | ||
| I implemented the HTF filter into my live system (v1.4.0) to see how it affected the broader metrics. |
- PF improved from 1.61 to 1.69.
- The Monte Carlo pass rate (MC), the probability of passing strict prop-firm rules based on resampled returns, rose from 94% to 95%.
- The M1 worst loss, which is the worst single-day drawdown calculated from 1-minute bar data, improved from 1.91% to 2.04%.
- Maximum drawdown (DD) improved slightly from -8.3% to -8.1%. However, there is a trade-off. Monthly returns dipped slightly from 0.77% to 0.74%, and the risk-adjusted efficiency remained effectively unchanged. In other words, adding the HTF filter is not a way to get “free” performance. Instead, you are trading a small amount of return for higher quality, lower drawdown, and better consistency. For those who prioritize regular withdrawals and account stability, this is a desirable shift. For others who prefer slightly higher raw returns, the current version of the system remains the better choice. Ultimately, the win rate for this strategy is structurally locked due to its trend-following nature, and no amount of filtering will push it significantly higher.
How this connects
This verification builds on earlier ones (what failed before and what I tried this time, comparisons between approaches).