Trade-level analysis, premium-path inspection, exit-policy comparison, and model evaluation for one capture run. Read alongside docs/ARCHITECTURE.md §2b before citing any number here.
Net P&L and trade count across every captured trade, grouped by instrument, option type, expiry, and entry hour.
All seven policies replayed on the untouched TEST trades only. The grey marker is what actually happened historically.
Favourable vs. adverse excursion during the trade window, and how much of the peak was actually captured at exit. MFE is historical opportunity, never a guaranteed price.
Predicted vs. realized forward return, feature importance, and performance against both baselines (see docs/ARCHITECTURE.md §2b on why "persistence" is the bar that matters).
Best and worst TEST trade by net P&L — premium path (with running peak) above, underlying spot below.
Aggregated across every trade's capture-time validation report.