Closed-loop intelligence: settled bets feed back into grading weights.
- Grade accuracy tracking per stat type (A/B/C/D hit rates)
- Signal accuracy tracking (which deltas predict outcomes)
- Kill condition effectiveness (hit_rate_with vs without)
- Conservative weight adjustment (20% cap, 50-pick minimum)
- 4 new DB tables: grade_accuracy, signal_accuracy,
kill_condition_accuracy, weight_history
- Desk-tier endpoints: /api/model/accuracy, /api/model/insights
Spec complete, ready to build when Phase 3 deployment is stable.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
builtbykev