Predicting International Success of Pace Bowlers in T20 Cricket

This study investigates the extent to which domestic T20 performance metrics can predict international success for pace bowlers in cricket. Using ball-by-ball data from over a decade of domestic and international T20 matches provided by the England and Wales Cricket Board (ECB), we engineer a comprehensive set of player-level features, including ball-tracking variables and outcome-based statistics. Success at the international level is evaluated using a Net Contribution metric adapted from the Duckworth-Lewis methodology. To identify key predictors, we apply feature selection techniques such as minimum redundancy maximum relevance (mRMR)andcorrelation clustering. Several regression models, including Random Forest and XGBoost, are trained and evaluated, with Random Forest achieving the best performance (R^2 = 0.53). Model interpretation using SHAP values reveals that a bowler’s boundary percentage, dot ball percentage and percentage of their wickets taken that were caught are among the most influential features. These findings offer data-driven insights for selectors and talent scouts seeking to identify and fast-track promising pace bowlers from domestic leagues.