A Hybrid Approach to T-20 Cricket Team Selection: Combining Probabilistic and Machine Learning Techniques

Abstract

This study investigates the use of advanced machine learning models to improve player selection in T-20 cricket, focusing on both batsmen and bowlers. Performance thresholds, based on first-quartile metrics such as bowling economy, strike rate, batting averages, and boundary-hitting ability, were used to identify top-performing players. Exploratory data analysis highlighted key relationships between performance indicators and selection decisions. The study evaluates the predictive accuracy of four machine learning models: Random Forest, Neural Networks, Logistic Regression, and Naive Bayes. Random Forest outperformed all other models, achieving perfect classification accuracy, while Neural Networks and Logistic Regression also showed strong results. Naive Bayes, a probabilistic model, demonstrated lower accuracy but provided valuable insights into performance patterns. These results show how machine learning and probabilistic models can help build stronger T-20 cricket teams by focusing on consistent and impactful players.