VWAP-Deviation Fade NET POSITIVE
535,002 trades · 53.7% win · gross +187068.92R · net +91892.65R · PF 1.88 (net 1.355) · maxDD −65.5R · P(loss) 0.0
IVA Break-In Fade NET POSITIVE
556,401 trades · 96.8% win · gross +194655.72R · net +93839.68R · PF 14.44 (net 3.508) · maxDD −8.5R · P(loss) 0.0
CVD-Divergence Fade NET NEGATIVE
109,236 trades · 43.3% win · gross +5147.60R · net -14872.62R · PF 1.09 (net 0.775) · maxDD −14890.8R · P(loss) 1.0
IVA Confirmed Breakout NET NEGATIVE
265,931 trades · 44.1% win · gross +16826.06R · net -31815.85R · PF 1.17 (net 0.749) · maxDD −31840.5R · P(loss) 0.991
Updated Analysis CORRECTIONS + FOLLOW-UP
Re-running E3/E4/E5/E8's own backtest code found a VWAP lookahead bug (E8), an entry-price lookahead bug (E4), a full-day volume-profile target lookahead bug (E3, found Sep 2026), a day-level ATR sizing bug shared by E3/E4/E5 (also found Sep 2026, fixed — moved every affected result by only a few hundredths of PF), and a shared crash bug. E5's implementation checked out clean. Corrected numbers, an out-of-sample-validated 15-variant follow-up study, a systematic OI+CVD+Volume confluence test, and a September 2026 hardening pass that closed out the two confluence categories which had survived the first correction (plus three further new signals, all null) are below.
Method
Discipline: one entry per signal (cooldown until flat), next-bar-open fills (taker), ATR-normalized R, net of taker+slippage. Strategies frozen before testing. Signal-only research, not investment advice. 600-day window covering the full checked-in dataset.