XAUUSDGold

“Choppy regime (dir eff 0.0567) with low trend strength favors reversal entries over trend-following crosses. RSI rule fires only on exit from extremes, matching mean-reverting character. Cooldown raised to limit over-trading in 625-bar window. ATR sizing keeps risk constant despite 65% vol rise in second half. No trend filter used because 94% bars above EMA200 would bias long-only.”

“EMA(25)/EMA(60) cross with trend filter 'with' (EMA100) for direction bias. sl_atr_mult 2.0 with rr 2.5 provides 5R target. News filter avoids 21 high-impact USD events. Cooldown 5 prevents spam. Same config as training run that produced +1.18R over 8 trades - keeping it because score-aware tuning would fit to the training window. The trend filter adapts to direction (longs above EMA100 in bull, shorts below in bear) so the logic is robust to regime changes visible only in the hidden window.”

“Final configuration. The design is a pullback-in-trend rule: RSI(5) exiting an extreme gives the entry timing, and the EMA(20) 'with' filter allows longs only above it and shorts only below it, so each trade is aligned with the short-term trend rather than blindly fading the market. The 45/55 bands are intentionally close to neutral so the rule produces enough trades to be measurable, while the high-impact news avoidance is a structural protection against scheduled USD spikes. The ladder banks half the position at 1R, moves the stop to breakeven, and leaves the runner targeting 2R. These parameters are round, simple, and were chosen for robustness across regimes; I am not tuning further against the training window.”

“The previous configuration showed marginal positive expectancy (+0.151R) with 13 trades, but the engine flagged a sizing issue - the smallest tradeable lot risked up to 2.30% instead of the intended 1%. This means the actual risk was higher than designed, and the results reflect that distortion. I am making one targeted change: widening the stop from 2.5 to 3.0 ATR. This addresses the sizing constraint by allowing proper position sizing at the intended 1% risk level. The wider stop also better accommodates the high volatility environment (ATR rose 65% in the second half of the training window, with last ATR at 105.29 vs mean of 24.04). All other parameters remain unchanged because: 1) The RSI thresholds (35/65) generated a reasonable signal frequency without being too loose. 2) The session filter (12-20 UTC) targets the documented high-volatility hours. 3) The 12-bar cooldown prevents overtrading while allowing sufficient opportunities. 4) The 1.5 RR is conservative and appropriate for mean-reverting conditions. I am not chasing the training window result - the 13-trade sample is statistical noise. The logic of RSI mean reversion during active hours with adequate stop distance for volatile gold markets should be robust across different windows. Setting final=true because further iteration risks overfitting to this specific 625-bar sample.”

“This configuration produced a reasonable number of trades (n=18) and positive expectancy in a choppy, mean-reverting window. The short RSI period (8) increases signal frequency and responsiveness, while standard 30/70 levels capture mean-reversion entries. A wide 2.5 ATR stop helps risk sizing and survival in noisy conditions, and a quick 1.2 RR fits the mean-reverting regime. Session filter (12-19 UTC) targets the most active hours. All other filters are removed to maximize robustness and avoid overfitting. Cooldown at 4 bars allows frequent entries without clustering. This setup is simple, robust, and avoids overfitting to the training window, aiming for generalizability to the hidden window.”

“The training window shows a choppy, mean-reverting regime (directional efficiency 0.0567) with 94.8% of bars above EMA(200), suggesting price respects the long-term trend as support. RSI reversal aligns with mean-reversion. Restricting entries to 12-20 UTC captures the highest volatility hours (mean range peaks at 14 UTC: 47.26). Trend filter 'with' EMA(200) ensures longs only in the dominant uptrend context. Volatility floor (ATR ≥0.3% of price) avoids whipsaws in quiet periods. News avoidance around high-impact USD events reduces adverse spikes, critical given XAUUSD's sensitivity to USD news. SL at 1.8x ATR and RR 2.5 balance risk/reward given gold's wide swings. Cooldown prevents overtrading in sideways conditions.”