Interpretable Prediction and Large-Scale Analysis of Judging in Professional Boxing

Authors: Mason duBoef, Thomas Romeas, Mathieu Charbonneau, and Allan Svejstrup

Professional boxing has long been plagued by judging controversy, driven largely by the sport's subjective scoring criteria. Leveraging new computer vision technology, we map round-by-round statistics to judges' scores at unprecedented scale (7,323 rounds). We introduce a points-based scoring system that achieves accuracy within the range of professional judges while remaining transparent, consistent, and bias-free. We also reveal which specific performance factors most strongly drive judges’ scores, offering actionable insights for improving scoring standards and competitive fairness in combat sports.