Football results are uncertain, but bookmakers express their expectations through betting odds. These odds can be studied as probability forecasts. This paper examines market-average closing odds for 1,140 Premier League matches from 2021/22 to 2023/24. Home, draw and away odds were converted into implied probabilities and proportionally normalized. The analysis was quantitative and reproducible, using publicly available closing odds and standard forecast evaluation measures. Forecasts were assessed using favourite accuracy, the multiclass Brier score, and five-bin calibration. The highest-probability outcome occurred in 58.25% of matches, and the Brier score was 0.551, compared with 0.667 for equal probabilities. Calibration was close overall, with an expected calibration error of 1.40 percentage points, although higher-probability outcomes occurred slightly more often than predicted. The odds provided useful forecasts and were well calibrated overall. However, assuming the normalized probabilities are correct, betting at these odds would still imply a theoretical loss due to the average overround of 3.91%.
No takes yet. Share an insight, caveat, or question.
Kening Lai (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: