Modeling Volleyball Match Outcomes Using Modified Estimators for the Binomial Parameter

Annual Meeting in Mathematics 2024

Abstract

Incorporating statistics into sports analytics is increasingly common, especially in predicting volleyball match outcomes. The discrete nature of volleyball scores, particularly the number of sets won, aligns well with binomial parameter estimators. While existing literature explores methods such as logistic regression, Bayesian approaches, and machine learning techniques, binomial estimation offers a simpler and more versatile approach that suits various scenarios in practice. This article introduces and compares six estimation methods for volleyball match prediction, all based on modified estimators of the binomial parameter. Simulations were conducted to forecast the outcome of the FIVB Volleyball Women’s Nations League 2023. Our findings indicate that the model using traditional Maximum Likelihood Estimation (TMLE) outperforms others, as evidenced by metrics such as Mean Square Error and Kendall’s Tau. While the TMLE demonstrates the highest predictive accuracy, it is important to recognize the advantages of models that adjust the estimation of the probability for tail values. Specifically, these models show a significant improvement in predictive capabilities for underdog teams, thereby enhancing the overall reliability in forecasting match outcomes.