Model calibration audit

For every bucket of predicted probability, what did the model actually hit? A well-calibrated model has actual win rate ≈ predicted prob. A negative gap means the model is overconfident in that bucket (dangerous); positive gap means it's underpredicting (safe). Rows turn red when |gap| > 5pp AND N ≥ 10.

Calibration cohort by sport (click to filter):
All sports MLB N=3,584TENNIS_WTA N=909TENNIS_ATP N=873MMA_MIXED_MARTIAL_ARTS N=388NFL_PRESEASON N=145LALIGA N=116NCAAF N=95SERIEA N=87EPL N=86LIGUE1 N=84UCL N=61BUNDESLIGA N=53NBA N=12NCAAB N=11NHL N=9AMERICANFOOTBALL_NFL N=2

Combined "all sports" is rarely meaningful — sports differ in market efficiency, signal availability, and base rates. Use the chips to drill into a single sport. N<50 (red) means the calibration is brittle; N≥200 (green) is trustworthy.

Filter: window=90d · sport=mma_mixed_martial_arts

Overall: N = 388 · mean predicted 59.6% · actual win rate 57.0% · gap -2.7pp · Brier 0.214 · log-loss 0.619
Calibration by predicted-probability bucket.
Predicted-prob bucket N Mean predicted Actual win rate Gap (actual − predicted) Brier
<50% 110 40.7% 34.5% -6.1pp 0.219
50-55% 54 52.4% 53.7% +1.3pp 0.248
55-60% 36 57.8% 44.4% -13.4pp 0.258
60-65% 44 62.1% 61.4% -0.7pp 0.240
65-70% 37 67.3% 73.0% +5.7pp 0.197
70-75% 32 71.9% 84.4% +12.4pp 0.151
75-80% 26 77.7% 69.2% -8.4pp 0.218
80-90% 36 83.0% 75.0% -8.0pp 0.195
90%+ 13 93.9% 92.3% -1.6pp 0.076