Adding a Third Strategy Cut Our Drawdown from 39.8% to 27.2% Without Touching the Other Two

When a system's drawdown is too deep, the reflex is to re-optimise the entry. We tried the other route on our XAUUSD system: leave both existing engines exactly as they were and add a third that decides differently.

The first is a trend core built on an ATR breakout with a regime filter. The second is a momentum breakout engine. The third, added in version 3.10, is an entry gate built with gradient boosting and validated by walk-forward analysis, so it is always judged on periods it did not learn from. All three run together in a single account, with no martingale, no averaging down, and a hard stop loss on every position.

Because the engines are close to uncorrelated, the combined system showed a shallower drawdown than the two-engine build: 27.2% maximum equity drawdown against 39.8%. The two original engines were not re-tuned to achieve that. The improvement came from combination, not from a better fit to the past.

Strategy tester, XAUUSD, 100% real ticks, initial deposit 10,000, leverage 1:500, January 2022 to July 2026, all three engines active. In Ultra mode: 35.1% annualised, profit factor 1.49, maximum equity drawdown 27.2%, over 533 trades. In Standard mode: 18.7% annualised, profit factor 1.45, maximum equity drawdown 18.5%. On a much longer window, 2004 to July 2026 with only the trend core and the momentum engine, the profit factor was 1.28 with a 44.4% maximum equity drawdown — a window that contains the 2008 crisis and the 2011 to 2015 gold bear market.

Each result was checked for rate invariance: the same profit rate has to reproduce on an independently rebuilt price feed, not only on one data source. Monte Carlo resampling gives a survival rate of 92%. Each engine, run standalone, reproduces the corresponding published product exactly, which confirms the integration did not quietly alter the logic. These are backtest results, not live trading results, and they do not indicate future performance.

If two of your strategies lose in the same weeks, adding a third of the same kind buys nothing. If it loses in different weeks, it lowers the worst case even when its standalone record is unremarkable. Check the overlap of the losing periods before you check the returns.

The system, with the full measurement history: https://www.mql5.com/en/market/product/187329

All products and free utilities: https://www.mql5.com/en/users/app.develop.sk/seller