Performance Assessment of the Machine Learning Algorithms for First Inning Score Prediction in Cricket

Abstract

This study presents a comprehensive performance assessment of various machine learning algorithms employed for the prediction of first inning scores in cricket matches. Cricket, a sport characterized by diverse formats like IPL, T20 Interna-tionals (T20Is), and One Day Internationals (ODIs), poses a unique challenge for score prediction due to its dynamic nature and format-specific nuances. We evaluate a range of machine learning models, including regression techniques, decision trees, and ensemble methods, across these cricket formats. Our analysis considers the significance of feature engineering, encompassing player statistics, pitch conditions, weather, and venue-specific factors, in enhancing predictive accuracy. Results indicate that the effectiveness of machine learning models varies across formats, with XGBoost regression performance in IPL and T20Is is better , whereas ridge regression performed well in ODIs. This research contributes valuable insights for cricket analysts, teams, and enthusiasts seeking to leverage machine learning for informed first inning score predictions, thereby enhancing their understanding of the intricate dynamics within the game.