Simulated execution of the copilot's own suggested 0DTE contract. v8 enters at ask and marks/exits at bid (60 s ticks); earlier cohorts remain mid-to-mid for historical continuity. v6+ rows run the management policy: spot touch of the cut exits, otherwise a premium floor ratchets with the running peak (−40% stop → breakeven at +30% → +25% at +50% → trail); T1 touch no longer exits. v2–v5 rows keep the first-touch model (cut or T1) they were graded under; still-open reads finalize at the close mark. Premium R = premium P&L / planned premium risk (capped at 40% of entry on v6+). Server-side over the full period — ignores the DAY TYPE filter. Tracking starts 2026-07-02 (engine v2).

The ranking question: higher confluence bands should perform better. This score is not a win probability; flat bars mean the ranking adds no edge. White tick = 50% coin flip.

R = 0.10% of spot (~7 SPX pts), same unit as the race grade. ≥NR = the move reached +N×R before first touching −1R (your stop). MFE→close runs to the bell, not 60 min; trend-day runs show up here. Premium translation on an ATM 0DTE (~0.5Δ): +2R ≈ +50–100%, +3R and beyond is 2x–10x territory. This table is the expectancy view for a runner style: low win rate is fine if ≥2R hits often enough.

Reads bucketed by confluence score. Look for monotonic outcome improvement; do not interpret the score as probability.

AGREE = Commander Read and copilot leaned the same direction at log time; DISAGREE = opposite directions; NO ENGINE READ = Commander was neutral/waiting (includes all reads before 2026-06-13, when engine bias wasn't yet recorded). Race 60m = the standard ±0.10% first-touch grade. Hold columns = if you'd held the bias for that long: win % is spot finishing in the bias direction, with average move in points.

Did the read keep working past the bell? Each column is the signed move (in the bias direction) from the read's spot to that session mark: NY+2h (18:00 ET), Asia (20:00), London (03:00 next morning), and the next NY open (09:30). Win % = share finishing in the bias direction; R-capture = reached +N×R (R = 0.10% of read spot) across the overnight path before −1R. Overnight marks are ES/NQ-proxied into the symbol's price scale (* = stale futures print). Fills the morning after each session. Respects the DAY TYPE filter.

Walls anchored at the first snapshot in the 09:30–10:00 ET window: the levels you'd sell against. TOUCHED = spot traded through the wall intraday (short strike tested); FAILED = closed beyond it (0DTE short at the wall finishes ITM); HELD = never touched. IC ✓ = closed inside both walls. Distance buckets matter: a wall 30+ pts out holding 90% is sellable, the same stat 8 pts out is not. Respects the DAY TYPE filter above.

Every copilot signal fired in the window — the same log as the Commander copilot box. Entry drives the cut-based R levels (1R/2R/3R price targets); Prem uses the cohort's recorded entry basis (v8 ask, earlier engines mid), and 30% is its +30% take. Exit fills once the premium tracker closes (else open). Respects the days/engine selectors; server-side.