A Performance Metrics–Based Model for Predicting Match Outcomes in the 2023 ICC Cricket World Cup

Background & objective: Match-outcome prediction in cricket requires the identification of performance indicators that distinguish winning from losing innings. This study aimed to identify key in-game performance indicators associated with match outcomes and develop a binary logistic regression model using data from the 2023 ICC Men’s Cricket World Cup. Materials and methods: A retrospective observational analytical design was employed. The dataset comprised 94 innings-level observations from 47 completed matches. One rain-affected match decided by the Duckworth–Lewis–Stern method was excluded because the second innings was truncated. Data were extracted from official International Cricket Council scorecards and the ESPNcricinfo statistical database. Predictors included toss outcome, opening partnership score, runs and wickets lost in the powerplay, total fours, total sixes, and total wickets lost in an innings. Binary logistic regression with forward stepwise variable selection was used to examine associations with winning and losing outcomes. Model fit was assessed using the Hosmer–Lemeshow test, and classification accuracy was evaluated on the development sample. Results: The final model retained two predictors. Total wickets lost in an innings was positively associated with the odds of losing (OR = 2.324, p < .001), whereas total fours was negatively associated with the odds of losing (OR = 0.863, p = .002). The model achieved an apparent classification accuracy of 81.9%. The Hosmer–Lemeshow test indicated no statistically significant evidence of poor fit (χ²(8) = 5.216, p = .734). These results identify wicket preservation and boundary scoring as complementary indicators associated with match outcomes. Conclusions: This study contributes tournament-specific evidence to international cricket-performance analysis by identifying two interpretable innings-level indicators associated with winning outcomes. However, the findings are limited to the tournament and analytical sample studied; dependence between innings, stepwise selection, and the absence of independent validation limit generalizability and predictive claims.