The influence of kinematic parameters of side-foot kick on classification of ball direction in football penalty kicks using machine learning and SHAP analysis techniques.

This study investigated how kinematic features could be used to predict soccer penalty kicks, using machine learning to improve goalkeeper anticipation. Fifteen right-foot-dominant male players participated in standardised assessments, with the goal area divided into six sections. Motion capture collected 3D kinematic data from 10 frames before support foot contact to ball strike. Class imbalance was addressed using the Synthetic Minority Oversampling Technique (SMOTE), and four machine learning classifiers, eXtreme Gradient Boosting (XGBoost), Random Forest (RF), Decision Tree (DT), and Artificial Neural Network (ANN), were trained to classify penalty kick direction. A Mutual Information-Genetic Algorithm (MI-GA) ranked and selected the top 49 most influential features. The left anterior superior iliac spine, left arm, and left humerus were key predictors, indicating that pelvic and upper-limb marker trajectories carried discriminative information for target-zone classification. XGBoost achieved the highest overall performance (accuracy = 0.821 ± 0.051), followed closely by RF (accuracy = 0.808 ± 0.051), whereas DT and ANN showed lower performance. Shapley Additive Explanations (SHAP) indicated that pelvic, humeral, and support-foot marker displacements contributed to classifying horizontal and vertical kick direction. These findings identify interpretable pre-contact kinematic cues associated with penalty-kick direction.