Shadow-outcome analysis and parameter recommendations (#904). Every bucket is a counterfactual replay — trades taken AND trades skipped, labeled under the strategy's locked stop/target/EOD logic.
| Bucket | n | Taken/Skipped | Hit rate | Avg R | Sum R |
|---|---|---|---|---|---|
| Loading… | |||||
| Skip cohort | Skips | Would-win rate | Avg R | R left on table |
|---|---|---|---|---|
| Loading… | ||||
"pattern_missing" = rejected because no pattern was detected at all; "low_confidence" = a pattern was found but scored below the threshold.
Joint tree-ensemble classifier over decision-time factors (vwap distance, RVOL, ATR%, gap%, float, ...), purged 5-fold CV vs. the raw AI-confidence baseline above. Applies only to AI Warrior Challenger strategies. Never produces a recommendation — see the AFML research track, EQUITY_LEARNING_SYSTEM.md §10.
No recommendations loaded yet.
Patterns that span your entire equity book — "bull flags on Fridays", "10-20% gap at open" — mined from counterfactual shadow outcomes (taken AND skipped), not just closed trades.
No global insights yet. Click "Mine Patterns" to run across your book.
Real-time warnings when a prospective entry matches a known weak bucket (block / warning)
or affirms a strong one (info). Driven by the Global Insights above.
No advisories yet. They accumulate as live equity signals fire; you can also use Preview above to test.
| Pattern | n | Hit rate | Avg R | Sum R |
|---|---|---|---|---|
| Loading… | ||||
| Band | n | Hit rate | Avg R | Sum R |
|---|---|---|---|---|
| Loading… | ||||
| Band | n | Hit rate | Avg R | Sum R |
|---|---|---|---|---|
| Loading… | ||||
| Band | n | Hit rate | Avg R | Sum R |
|---|---|---|---|---|
| Loading… | ||||
| Session | n | Hit rate | Avg R | Sum R |
|---|---|---|---|---|
| Loading… | ||||