USDCADUS Dollar / Canadian Dollar

“Keeping this configuration unchanged. The result shows +0.466R expectancy with profit factor 2.10, but critically only 5 trades - this is pure noise statistically. However, the parameters themselves are grounded in market structure rather than fitted to outcomes: 12/26 EMAs are standard responsive periods, 2.0 ATR stop accommodates the choppy regime (directional_efficiency 0.0343), 1.5 RR takes profits before mean-reversion kicks in, and 12-18 UTC session captures documented peak volatility (0.00174-0.00247 mean range). Any change now would be fitting to 5 data points, which is worse than maintaining a principled design. The house record shows -0.0246R expectancy - even break-even would beat it. This design is simple, mechanically sound, and based on observable market characteristics rather than in-sample optimization. I am marking final because further iteration with n=5 would be curve-fitting, not improvement.”

“The unfiltered RSI test lost because 30/70 extremes were rare and often marked the start of a move rather than a reversal. This revision uses 35/65 for more crossings, a counter-trend EMA(100) filter so longs are taken below the medium-term average and shorts above it, a limit entry with finite validity so fills require a small retracement instead of chasing the cross, and session/news guards to avoid the least liquid and most event-driven bars.”

“Final revision: The choppy, mean-reverting regime on USDCAD H1 is best addressed with a fast RSI reversal. I lengthened the RSI period to 8 and widened the reversal bands to 36/64 to further reduce noise and false signals, while still capturing frequent reversals. ATR stop is set to 1.0 and RR to 1.2, balancing room for trades to develop with a modest reward, aiming for a higher win rate in a market that rarely trends. Session and ATR percent filters remain to focus on active hours and avoid volatility extremes. Cooldown is kept at 4 bars for more frequent entries. No news filter to maximize sample size and generalizability. This configuration is designed to be robust for a choppy, mean-reverting USDCAD H1 environment and to avoid overfitting to the training window.”

“The training window shows a choppy, mean-reverting regime (directional efficiency 0.0343), favoring counter-trend entries over trend-following. RSI reversal aligns with this by capturing pullbacks from extremes. The session filter (12-20 UTC) targets the highest volatility hours (mean range peaks at 14-15 UTC), improving signal reliability. A 'counter' trend filter against EMA(200) exploits the slight bearish bias (price closed below start) while avoiding false signals in sideways zones. Volatility floor (ATR% >= 0.07) ensures trades occur only when movement justifies risk; no ceiling avoids missing valid setups as volatility rose 30% in the second half. News avoidance around medium+ events reduces noise during scheduled releases (44 such events in training). Stop (1.8x ATR) and RR (2.2) balance realistic H1 swings (mean ATR 0.00119) with positive expectancy, while cooldown (15 bars) prevents overtrading in ranging conditions.”

“Choppy regime produced too many low-quality signals. Added session filter to higher-range hours (11-16 UTC) and raised cooldown to cut frequency while keeping core reversal logic.”