Construction of 2022 Qatar World Cup match result prediction model and analysis of performance indicators

Abstract

This research investigates the influence of performance metrics on match outcomes and constructs a predictive model using data from the Qatar World Cup. Employing magnitude-based decision and an array of machine learning algorithms, such as Decision Trees, Logistic Regression, Support Vector Machines, AdaBoost, Random Forests, and Artificial Neural Network, we examined data from 59 matches, excluding extra time. Fourteen performance indicators were integrated into the model, with two types of match outcomes—winning and non-winning—serving as the output variables. The ANN model exhibited the highest predictive performance, achieving an accuracy of 75.42%, an AUC of 76.96%, a precision of 72.73%, a recall of 65.31%, a specificity of 77.03%, and an F1 score of 68.82%. SHAP analysis revealed that “On Target”, “Shooting Opportunity”, and “Ball Progressions” were the most influential features. These findings underscore the critical role of shooting accuracy and the creation of scoring opportunities in determining match outcomes. Consequently, this study developed an accurate model for predicting match outcomes and meticulously analyzed the match performance. Coaches should prioritize the sensitive indicators identified in this study during training and structure training sessions accordingly.