Live OOS Scoreboard
Per-detector hit-rate, Brier score, and reliability calibration on real recorded outcomes from the production `signal_outcomes` ledger — not synthetic walk-forward fixtures.
Outcomes recorded
52,829
Hit rate
23.0%
Brier score
0.5284
n=52829 (rows w/ confidence)
Earliest sample
11/20/2025, 6:19:01 PM
generated 10/9/2026, 6:41:10 PM
Per-detector scoreboard
Sorted by Brier ascending (lowest = best calibrated). Detectors with no confidence-bearing rows fall to the bottom by hit-rate.
| Detector | N | Hit rate | Brier | Brier N |
|---|---|---|---|---|
| falling wedge | 1619 | 7.4% | 0.2392 | 1619 |
| bull pennant | 100 | 17.0% | 0.2500 | 100 |
| cup with handle | 60 | 58.3% | 0.2528 | 60 |
| downtrend line breakout | 93 | 48.4% | 0.2555 | 93 |
| bear flag | 117 | 0.0% | 0.2589 | 117 |
| triple bottom | 344 | 34.3% | 0.2717 | 344 |
| ascending triangle | 620 | 29.4% | 0.2732 | 620 |
| symmetrical triangle | 57 | 57.9% | 0.3004 | 57 |
| inverse head shoulders | 530 | 33.0% | 0.3059 | 530 |
| double bottom | 119 | 30.3% | 0.3344 | 119 |
| pre breakout coil | 1999 | 16.2% | 0.3468 | 1999 |
| rectangle breakout | 376 | 21.5% | 0.3718 | 376 |
| channel breakout | 1387 | 25.9% | 0.3741 | 1387 |
| rectangle | 60 | 43.3% | 0.3746 | 60 |
| double top | 1611 | 12.2% | 0.3862 | 1611 |
| head shoulders | 7477 | 5.6% | 0.4174 | 7477 |
| bull flag | 65 | 27.7% | 0.4202 | 65 |
| rising wedge | 195 | 16.4% | 0.4298 | 195 |
| triple top | 1428 | 6.4% | 0.4320 | 1428 |
| bear pennant | 27 | 0.0% | 0.4351 | 27 |
| cup handle | 15 | 0.0% | 0.4392 | 15 |
| descending triangle breakdown | 1873 | 6.7% | 0.4486 | 1873 |
| ma bounce 200 | 542 | 53.1% | 0.4628 | 542 |
| ma reclaim | 60 | 16.7% | 0.4954 | 60 |
| breakout retest hold | 28292 | 31.2% | 0.6104 | 28292 |
| rounded base | 3602 | 16.1% | 0.7092 | 3602 |
| ma bounce 50 | 153 | 8.5% | 0.8706 | 153 |
| squeeze breakout | 8 | 0.0% | 0.9801 | 8 |
Reliability calibration
Each dot is a confidence-decile bucket: x = mean predicted probability, y = realized hit-rate. Dashed diagonal is perfect calibration. Dot size scales with bucket count.
| Decile | Range | N | Predicted | Actual | Gap |
|---|---|---|---|---|---|
| 1 | 0.0%–10.0% | 0 | – | – | – |
| 2 | 10.0%–20.0% | 0 | – | – | – |
| 3 | 20.0%–30.0% | 0 | – | – | – |
| 4 | 30.0%–40.0% | 119 | 35.7% | 12.6% | -23.1% |
| 5 | 40.0%–50.0% | 1700 | 46.0% | 13.5% | -32.5% |
| 6 | 50.0%–60.0% | 4512 | 55.4% | 16.7% | -38.7% |
| 7 | 60.0%–70.0% | 12316 | 65.2% | 15.2% | -50.0% |
| 8 | 70.0%–80.0% | 6432 | 72.7% | 36.0% | -36.7% |
| 9 | 80.0%–90.0% | 1448 | 82.4% | 41.2% | -41.3% |
| 10 | 90.0%–100.0% | 26302 | 96.5% | 24.2% | -72.3% |