NYCE ML tracker · external sports-betting-mcp MLB benchmark · opinion only
Clean ledger artifact · 365d · same-day rows are flagged until first-pitch snapshots are added
X-for-Y model artifact · feature store · no-leak serving status
Powered by PropLine · Lines move; this is what actually happened. Green = over 50% hit rate.
| Outcome | Date | Engine | Pick | Type | Side | Model % | Confidence | Line | Actual | |
|---|---|---|---|---|---|---|---|---|---|---|
| Loading picks… | ||||||||||
| Day | Date | Record | Win % | Total | Pending | Streak |
|---|---|---|---|---|---|---|
| Loading… | ||||||
Which prop categories are producing real value (not just wins).
| Prop Type | Record | Hit Rate | ROI | Units | Avg Odds | Picks | Status |
|---|---|---|---|---|---|---|---|
| Loading… | |||||||
How well does the model's predicted probability match actual win rate? A perfectly calibrated model's bars should match the dashed 45° line.
Trained ML models, ranked by holdout performance. Each was validated on a held-out season and only ships to the live blend if it beats the baselines. Expand a model to see calibration, lift, recall, and what it's blind to.