Track Record
Live model performance — every prediction logged vs actual outcome.
85-28
Record (all picks)
75.2% win rate
52-5
Sharp Record
91.2% on Sharp picks
69.0%
Avg Confidence (Sharp)
model's stated win probability
+11.4%
30D ROI (Sharp)
$100 flat, n=6 tracked
Rolling Log-Loss (10-game window)
Lower is better. Random baseline = 0.693. Well-calibrated model ≈ 0.64.
Cumulative ROI
$100 flat bets, 113 of 113 picks with a tracked opening line.
Calibration Curve
Dots on diagonal = perfectly calibrated. Size = sample count.
Win Rate by Predicted Confidence Bucket
When we say 70%+, are we actually right 70% of the time?