Your Calibration Track Record
The /audit page measures the engine's calibration. This page measures yours. Every pre-mortem you record (Ctrl/Cmd+Shift+P) captures your stated confidence; at horizon-end you resolve it. Patterns surface here — over time we learn whether your gut is calibrated, where it wins, where it doesn't.
Headline
Calibration curve — your confidence vs reality
Each bar is a confidence bucket (0-10%, 10-20%, …). Bar height = realized hit rate. Dashed line = perfect calibration. Bars below the line = overconfident; above = underconfident.
By engine tier
Where your confidence lines up best/worst with reality.
| Tier | n | Hit rate | Mean conf. | Bias |
|---|
By asset class
Where your judgment is calibrated, where it isn't.
| Class | n | Hit rate | Mean conf. | Bias |
|---|
You vs the engine
Across decisions where you and the engine had a stated probability, who's been right more often? Use disagreements as research prompts; defer to whichever historically wins per cohort.
| Tier | n | Agreement | You only right | Engine only right |
|---|