HHAD dual-boundary v8 locked evaluation
M3 supplies ordered probabilities; draw and cover gates only change the audited decision.
| market | model | status | samples | log_loss | brier | rps | hard_hit | draw_precision | draw_recall | predicted_draws | actual_draws | letwin_false_rate | bootstrap_improvement_probability |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| HHAD | M1_joint | BASELINE | 7 | 0.9294760449106777 | 0.5433903884095906 | 0.3353633835606208 | 0.5714285714285714 | 0.0 | 0.0 | 0 | 2 | 0.3333333333333333 | |
| HHAD | M5-HHAD-DUAL-BOUNDARY | REJECTED | 7 | 1.0248796072699906 | 0.6279960521526105 | 0.4025727683966431 | 0.2857142857142857 | 0.0 | 0.0 | 2 | 2 | 0.0 | 0.118 |