Predicting Cricket Outcomes using Bayesian Priors

TL;DR

This paper introduces a Bayesian statistical model to predict cricket tournament outcomes by incorporating players' historical performance and team opposition data, validated on IPL 2020 and applied to ICC World Cup 2023 predictions.

Contribution

The paper presents a novel Bayesian modeling approach that integrates stratified survey sampling and performance history for cricket outcome prediction.

Findings

Accurately predicted IPL 2020 top three teams including the winner.

Predicted probabilities of winning for all teams in the ICC World Cup 2023.

Model demonstrated reasonable accuracy and can be extended to other tournaments.

Abstract

This research has developed a statistical modeling procedure to predict outcomes of future cricket tournaments. Proposed model provides an insight into the application of stratified survey sampling to the team selection pattern by incorporating individual players' performance history coupled with Bayesian priors not only against a particular opposition but also against any cricket playing nation - full member of International Cricket Council (ICC). A case study for the next ICC cricket world cup 2023 in India is provided, predictions are obtained for all participating teams against one another, and simulation results are discussed. The proposed statistical model is tested on 2020 Indian Premier League (IPL) season. The model predicted the top three finishers of IPL 2020 correctly, including the winners of the tournament, Mumbai Indians, and other positions with reasonable accuracy. The method can predict probabilities of winning for each participating team. This method can be extended to other cricket tournaments as well.