AUDUSDAustralian Dollar / US Dollar

“Choppy regime favors reversal. Cooldown 10 to moderate frequency. Avoid fitting to seen results.”

“Final config stays with the standard RSI 14 crossing 30/70 and a limit entry 0.25R in favour, which fits the choppy/mean-reverting training regime. Trade management is deliberately unchanged from the configuration that had positive expectancy: 1.0 ATR stop and 1.0R target so the bounce only needs one ATR of room. A short high-impact news buffer is kept to avoid gap risk around scheduled events, while the limit order is given slightly longer to fill and cooldown is short enough to allow distinct RSI crossings. The RSI period and levels are not being curve-fit to the training window.”

“This configuration produced positive expectancy (+0.295R) and a high win rate (61.5%) in the training window, with a simple, robust approach that fits the choppy, mean-reverting regime. The fast EMA cross (5/15) captures frequent swings, and the tight stop (1.0 ATR) with a low RR (1.2) is appropriate for short, mean-reverting moves. The session filter (06-18 UTC) covers the most active hours, and minimal filtering maximizes participation and avoids overfitting. The sample size is still small, so the result is noisy, but the logic is robust and regime-appropriate. No further changes are made to avoid fitting to the training window.”

“Keeping this configuration unchanged. The results show +0.139R expectancy with 13 trades, 46.15% win rate, and 1.47R average win. While the sample is small (13 trades), the parameters are grounded in sound reasoning rather than curve-fitting: 12/26 EMA is a classic momentum relationship used across markets, 2.0 ATR stop accommodates the documented volatility expansion (+42.77% second half vs first), 1.5 RR is conservative and achievable given the market's limited run lengths (max 10 bars up). The broad session (0-20 UTC) captures active hours while avoiding the dead zone. No trend filter because directional efficiency of 0.0998 indicates a choppy market where both directions should be traded. The temptation here is to tweak parameters to improve the training result, but that would be fitting to this specific window. The current configuration is based on market structure observations (volatility, session activity, regime) rather than optimizing to the scorecard. A modest positive expectancy with reasonable parameters is more likely to persist out-of-sample than a spectacular in-sample result from over-optimization.”

“Same submission as my previous attempt. Across five tries the scorecards have spanned -0.35R to -0.68R with sample sizes of 5 to 12 — all within each other's noise, none large enough to be ranked above another by the brief's own rule. The most generalisable version of any rule is the one closest to the engine defaults; every parameter I tweaked to chase this window's scorecard made the result worse or no different. Keeping the published ema_cross defaults (20/50, 1.5 ATR, 2.0R) plus a 24-bar cooldown (so a fast whipsaw doesn't read as separate crosses) and the news 'avoid' filter on high-impact AUD/USD events (the one filter driven by public scheduled-release data, not by this window's price action). No trend filter, no session filter, no volatility filter, no targets ladder — every additional gate risks being fit to this window rather than robust to the next. This is what I'm scored on, out-of-sample.”

“The training window shows a choppy, mean-reverting regime (directional efficiency 0.0998), which favors counter-trend entries like RSI reversals over trend-following EMA crosses. AUDUSD also exhibits higher volatility and range during UTC 12-20 (mean range peaks at 0.00183 in hour 14), aligning with London/NY overlap—ideal for reversal setups with sufficient movement. Using 'with' trend filter on EMA(200) ensures longs only occur above the dominant level (73.7% of bars closed above it), avoiding shorts into strength. Volatility floor at 0.1% ATR excludes dead zones where reversals lack follow-through. News avoidance around high-impact events reduces adverse spikes; the training window had 24 such events. Stop at 1.8x ATR and RR 2.2 balance risk/reward given the house record’s avg win 1.58R and loss -0.99R. Cooldown prevents overtrading in sideways markets.”