Smash landing in elite male badminton players: Effects of fatigue and task predictability via explainable machine-learning analysis

Smash landings in badminton carry high injury risk, worsened by fatigue. Assess how fatigue and task predictability shape lower-limb biomechanics in elite male players and identify fatigue markers via explainable ML. Fourteen elites performed anticipated and unanticipated smash-landings before and after fatigue. Synchronous 3-D kinematics/kinetics were analyzed with a 2 × 2 repeated-measures ANOVA. XGBoost models based on anticipated-task trials were used to classify fatigue status; SHAP ranked feature importance. Fatigue and unpredictability independently affected peak coronal-plane knee moment (fatigue p = 0.033; task p = 0.016). Significant ankle-moment effects were also observed in the transverse plane. Statistical parametric mapping highlighted time-specific effects of fatigue and task on knee/ankle sagittal moments, with an interaction for knee sagittal ROM ( p = 0.0395). XGBoost showed high accuracy, identifying peak ankle plantar-flexion moment as the foremost fatigue discriminator. Fatigue and unanticipated tasks were associated with reduced lower-limb control during smash landings. Knee moment and vertical ground-reaction force may serve as biomechanical indicators of fatigue-related movement alterations. Training strategies that combine fatigue exposure with decision-making demands may help improve motor control under high-demand task conditions.