Consistency of Judging Under the World Boxing 10-Point Must Scoring System: Single-Judge and Five-Judge Aggregate Reliability Across the 60 kg, 75 kg, and 90+ kg Categories at the 2025 World Boxing Championships

Background: Reliable judging is essential to the fairness of elite boxing, yet inter-judge reliability under the current World Boxing 10-Point Must Scoring System remains largely unexplored. This study aimed to evaluate inter-judge reliability across rounds and weight categories at the 2025 World Boxing Championships. Methods: A retrospective observational analysis included 94 men’s bouts in the 60 kg (n = 42), 75 kg (n = 23), and 90+ kg (n = 29) categories. Judge-level round scores were analysed using linear mixed-effects variance-component models, with bout-round target and judge identity specified as random effects. Reliability was estimated for individual judge ratings (Rsingle) and an aggregate measurement based on five judge ratings (Rpanel(5)). Pairwise differences across rounds and weight categories were evaluated using bout-level bootstrap procedures with Holm adjustment for multiple testing. Results: Across all bouts, Rsingle was 0.617 (95% CI: 0.521–0.693) in Round 1, 0.634 (95% CI: 0.539–0.709) in Round 2, and 0.687 (95% CI: 0.597–0.754) in Round 3. The corresponding Rpanel(5) estimates were consistently higher at 0.889 (95% CI: 0.845–0.919), 0.897 (95% CI: 0.854–0.924), and 0.916 (95% CI: 0.881–0.939), respectively. No statistically significant differences in reliability were detected across rounds or between weight categories after Holm adjustment. Overall, 98.2% of judge-round scores were 10:9 and 68.9% of bouts decided by judges’ scores ended in unanimous 5:0 decisions. Conclusions: Reliability was consistently higher when five judge ratings were aggregated than when individual judge ratings were considered. No statistically significant evidence of systematic differences in reliability across rounds or between the examined weight categories was detected.