A Counterfactual Sensitivity-Based Learning Framework for Outcome Prediction in One-Day International Cricket

Predicting outcomes in One-Day International (ODI) cricket is difficult due to rapidly changing match conditions and complex in-game dynamics. Traditional machine learning models often rely on static representations and fail to capture the sensitivity of outcomes to realistic variations in matches. This paper proposes a Scenario-Aware Counterfactual Sensitivity-Based Learning (SCIL) framework for predicting ODI match outcomes. The framework generates plausible counterfactual match scenarios and assigns influence weights based on prediction sensitivity, enabling the model to emphasize decisive match situations. Experiments on historical ball-by-ball ODI data show that SCIL consistently outperforms conventional models, including Logistic Regression, Random Forest, and XGBoost. The results show enhanced robustness, predictive accuracy, and probabilistic discrimination, highlighting the effectiveness of counterfactual sensitivity modeling for reliable cricket match prediction outcomes.