Calibration
Postmortem.
Every signal is checked against what actually happened — did the price move in the direction the thesis predicted? Pillars whose reads consistently prove out earn more weight in future scoring; those that don't get a higher bar to clear before they influence decisions.
| Pillar | Threshold | 60d accuracy | Reads | Updated |
|---|---|---|---|---|
| P1 | 0.70 | — | 0 | 2026-05-04 |
| P2 | 0.70 | — | 0 | 2026-05-04 |
| P3 | 0.70 | — | 0 | 2026-05-04 |
| P4 | 0.70 | — | 0 | 2026-05-04 |
| P5 | 0.70 | — | 0 | 2026-05-04 |
Pillars with fewer than 20 scored reads keep the default threshold. Thresholds recalibrate on the 1st of each month.
Top false positives · 30d
None yet.
Top missed-opportunity tickers · 30d
None yet.
TP
0
FP
17
Neutral hit
83
Neutral miss
0
Insufficient
900
Each signal is evaluated at three horizons: 5 days, 20 days, and 90 days. The 5-day window drives threshold recalibration.