Understanding the effect of contextual factors and decision making on team performance in Twenty20 cricket: an interpretable machine learning approach

For better performance in any team sport, team managers assess the match conditions, and opponents’ strengths and weaknesses to select the best team possible. In cricket, existing studies focus on the effect of contextual factors such as home advantage, toss win, and toss decision, among others, on team performance. However, very few studies discuss the factors’ relative importance or the extent of their impact on performance. There is also a lack of studies addressing the best situational decisions to be taken by teams, given certain opponents and match conditions. This study aims to determine the effect of contextual factors and subsequent decisions taken on team performance in Twenty20 (T20) cricket. Match-wise data for nine seasons of the Indian premier league consisting of 563 matches were considered, and tree-based machine learning (ML) models such as gradient boosting, regression tree, bagging, random forest, and bayesian additive regression tree (BART) were employed for data analysis. BART produced the most efficient results, which were further interpreted using Interpretable ML methods such as partial dependence plots and accumulated local effects to determine the most critical factors affecting team performance. Additionally, these findings were used to obtain optimal pre-match decisions and pre-season strategies to achieve higher performance, which could serve as a decision support system for teams in T20 cricket.