A risk-off signal that helped the FX core and nothing else
Filtering for risk-off signals based on price momentum does not reduce the overall system drawdown, even though it improves the performance of the…
Filtering for risk-off signals based on price momentum does not reduce the overall system drawdown, even though it improves the performance of the core FX component. I recently tested whether a directional, price-based signal could outperform my current filter of using the US500 index below its 200-day simple moving average (SMA). I experimented with faster SMAs (100 and 50), negative momentum, volatility spikes, and multi-index breadth.
FX Core Performance
When isolating the FX core component, faster SMA triggers showed a clear, monotonic improvement in risk management.
| Metric | SMA 200 (Baseline) | SMA 100 | SMA 50 |
|---|---|---|---|
| Max Drawdown | -7.4% | -6.9% | -6.5% |
| Return/DD Ratio | 0.24 | 0.25 | 0.27 |
| In other words, the faster SMA triggers allowed the system to identify market stress earlier. While this performed well during crisis periods like 2015-2018 and 2024-2026, it suffered during strong trend periods (2021-2024), resulting in a success rate of 2 out of 4 sub-periods. |
Full System Impact
Despite the improvement in the core, the full system v1.4.1 failed to see a reduction in total drawdown.
| Metric | SMA 200 (Baseline) | SMA 50 |
|---|---|---|
| Max Drawdown | -9.6% | -9.7% |
| Monthly Return | 0.84% | 0.84% |
| Profit Factor | 1.67 | 1.67 |
| MC Pass Rate | 94% | 94% |
| Using the SMA 100 actually worsened the drawdown to -10.3%. This confirms a recurring lesson from my research: improving a single component does not necessarily improve the system. The total drawdown is dictated by the correlation between different sleeves, not just the performance of one. Even if I time the FX core’s risk-off exit more aggressively, the system remains constrained by its structural dependencies. |
Practical Takeaway
The only meaningful benefit of the SMA 50 filter was a reduction in the “M1 worst” drawdown from 2.42% to 1.69%. This metric tracks the worst intraday loss using 1-minute bars. Because the SMA 50 triggers an earlier exit, it successfully suppresses large intraday losses. I am keeping v1.4.1 as-is with the SMA 200 filter, but I will keep the SMA 50 logic in my toolkit as a “safety valve” for high-leverage scenarios, similar to a scale-out strategy. Ultimately, to lower the system drawdown, I need truly uncorrelated, positive expected value sleeves, which I have yet to find within price-based data alone. This reinforces that v1.4.1 is currently sitting on the optimal drawdown frontier.
How this connects
This verification builds on earlier ones (what failed before and what I tried this time, comparisons between approaches).