59%
ELO baseline: 55%
Accuracy
0.642
Random = 0.500
AUC
255
of 1273 total
Test Fights
LR
Logistic
Model
Calibration — Predicted vs Actual Win Rate
0-20%
12%→15%(13)
20-40%
32%→44%(68)
40-60%
49%→52%(102)
60-80%
68%→71%(52)
80-100%
93%→70%(20)
Purple = predicted · Coloured bar = actual · Green = well-calibrated
Confidence:
Showing 267 fights61% accuracy (163 correct)
Model: Logistic Regression (pure Python fallback) · Trained 2026-05-21 · 1018 training fights