Active validation stagePAPER VALIDATIONReal-money staking stays disabled until calibration and CLV thresholds hold across a full season.
Odds API uptime
96.9%
OK
Kalshi uptime
85.1%
OK
Storage uptime
100.0%
OK
Degraded / failed runs
76
Last run —
Stale-data alerts
172
Runs with odds older than 900s
Zero-liquidity rejections
176
Guardrail blocked untradeable markets
Market-match preventions
15
Incorrect market pairings blocked pre-bet
Brier score
0.230
Lower is better · 0.25 = coin flip
Rejection reasons
Why candidates were passed or declined.
- Edge below threshold478
- Bookmaker count below minimum (< 4)335
- Zero liquidity on Kalshi market176
- Commence time inside blackout window17
- Stale odds (> 900s)16
- Market match confidence too low15
- Consensus dispersion too high13
Bookmaker count distribution
How many books priced each candidate. Fewer than 4 is auto-rejected.
Consensus dispersion
Standard deviation across book prices. High dispersion means an unreliable consensus.
Model calibration
Predicted probability vs. observed win rate per bucket.
| Bucket | Predicted | Observed | N | Gap |
|---|---|---|---|---|
| 0–10% | 5% | 0% | 2 | -5% |
| 10–20% | 15% | 27% | 11 | +12% |
| 20–30% | 25% | 23% | 13 | -2% |
| 30–40% | 35% | 49% | 49 | +14% |
| 40–50% | 45% | 54% | 71 | +9% |
| 50–60% | 55% | 64% | 83 | +9% |
| 60–70% | 65% | 72% | 95 | +7% |
| 70–80% | 75% | 56% | 45 | -19% |
| 80–90% | 85% | 87% | 15 | +2% |
| 90–100% | 95% | 80% | 10 | -15% |
Stale-data alerts
Runs where the odds snapshot exceeded the 900-second freshness budget.
- run_00111127s oldSYSTEM OK
- run_00321238s oldSYSTEM OK
- run_00401339s oldSYSTEM DEGRADED
- run_0052961s oldSYSTEM OK
- run_0072956s oldSYSTEM OK
- run_00821389s oldSYSTEM OK
- run_00901369s oldSYSTEM OK
- run_01111273s oldSYSTEM OK
- run_01211249s oldSYSTEM OK
- run_01401253s oldSYSTEM OK
- run_0151902s oldSYSTEM OK
- run_0161901s oldSYSTEM OK