What happens when you chase settings that won recently
In my recent research (Study 91), I tested whether a 4% monthly return with a maximum drawdown of 10% is achievable.
Finding the ceiling for monthly returns
In my recent research (Study 91), I tested whether a 4% monthly return with a maximum drawdown of 10% is achievable. I evaluated four different approaches, and the conclusion is clear: it is currently impossible. To reach a 4% monthly return while keeping the drawdown within 10%, a system would need a Calmar ratio (annualized return divided by maximum drawdown) of 4.8. My best-performing systems hover around 2.5 during favorable regimes and drop to 1.4 over full market cycles. I refuse to fabricate a system to hit an arbitrary target, so I am sticking to the reality that 0.7% to 1.8% is the current performance ceiling for safe, risk-managed trading.
| Strategy Approach | Monthly Return | Drawdown |
|---|---|---|
| v1.3.0 (Leverage increase) | 2.33% | -40% |
| Recent Market Basket | ~4.0% | -22% |
| Per-pair Specialized | ~4.0% | -18 to -20% |
| Vol-target Adjusted | No improvement | - |
The case for per-pair specialization
In Study 92, I moved away from a “one-size-fits-all” logic and built a system that assigns a specific, optimized archetype to each currency pair. This approach is significantly more effective than my previous uniform systems. By training the best logic for each asset (such as ATR-based candles for Gold and CADJPY, Donchian channels for USDJPY, and Connors RSI for indices), I created a basket of nine pairs that thrives in the current market regime.
- Performance (OOS 2022-26): +1.78% monthly return with a -7.8% drawdown.
- Risk Metrics: Sharpe ratio of 2.46 and a Monte Carlo (MC) pass rate of 97%.
- Safety: Zero days of M1 intraday threshold violations using a risk setting of 0.0025.
In other words, this specialized system provides about 2.5 times the monthly return of my Core v1.3 system (which averages 0.7%) while maintaining a similar risk profile. I have formalized this as the
perpair_system.
Important warnings on robustness
While the per-pair system is a powerful tool, it is not a “set and forget” solution. Its performance is heavily driven by the current regime of Gold strength and Yen weakness. If we look at the full data period from 2015 to 2026, which includes weaker regimes, the drawdown on this system expands to -14.9%, exceeding my 10% limit. It is a system built to lean into the current market, not one designed for absolute long-term robustness. For long-term capital preservation, my Core v1.3 system remains the primary choice. I now view the per-pair system as an aggressive secondary option that requires active regime monitoring and annual re-selection of pairs and archetypes.
Why Gold alone is a trap
Many have asked if focusing exclusively on Gold is the answer, given its high profit factor (PF). My verification shows that the impressive PF of 1.70 for Gold in the per-pair OOS test is heavily skewed by the 2024-2026 period. When you look at cleaner data from 2015 to 2024, the PF drops to 1.26, the monthly return falls to 0.66%, and the drawdown balloons to -18.5%. Because Gold lacks diversification, concentrating on it alone forces you to accept huge drawdowns. If you restrict Gold to a 10% drawdown, your monthly return drops to roughly 0.4%, which underperforms both my Core and per-pair systems. Gold is a powerful component of a basket, but it is not a standalone solution.
How this connects
This verification builds on earlier ones (what failed before and what I tried this time, comparisons between approaches).