Decision-making performance and motor skill consistency in basketball short pass–shot sequences under different levels of mental fatigue: an integrated psychophysiological dynamics analysis

Figure 1. Integrated data acquisition and analytical pipeline for the basketball short pass-shot sequence. The workflow strictly outlines the synchronized raw signal inputs, event-locked preprocessing, feature extraction, and the final inferential and predictive modeling stages employed in the study. (A) Inputs and Synchronized Acquisition. (B) Event-Locked Preprocessing. (C) Feature Engineering. (D) Integrated Statistical and Predictive Modeling.

Figure 2. Task-specific sensor deployment and event timing alignment for the short-pass-shot sequence under mental fatigue. (A) Sensor placement and coordinate system definition for the wrist-worn IMU sensor. (B) Basketball court diagram illustrating the pass and shot scenario with sensor placements on the human figure. (C) Trial timeline and event-locked data windows for the pass, catch, and shot segments.

Figure 3. Fatigue-induced alterations in decision-making and skill execution. (A–E) depict condition-wise comparisons for decision accuracy, decision latency, sequence success rate, pass landing error, and catch-to-shot timing variability, respectively. Data are presented as individual trial-aggregated data points with superimposed mean ± SD. Brackets indicate exact post-hoc significance levels for between-condition differences.

Figure 4. Fatigue-dependent variations in movement kinematics during the pass-catch-shot sequence. (A) Overlaid density-normalized histograms illustrating the broadened distribution of resultant acceleration magnitude as mental fatigue increases. (B) Boxplots comparing resultant acceleration magnitude across the three fatigue conditions, with individual data points and outliers depicted to summarize inter-condition dispersion.

Figure 5. Comprehensive cross-domain correlation matrix illustrating the coupling among psychophysiological markers, decision-making metrics, and skill-consistency indices. To accurately capture generalized functional relationships rather than transient trial-level noise, Pearson correlation coefficients were computed using participant-level summaries aggregated across all three experimental conditions (N = 27).

Figure 6. Machine learning classification matrices for fatigue-induced performance states. (A) Confusion matrix for decision-quality classification utilizing a Random Forest model. (B) Confusion matrix for skill-consistency degradation classification utilizing a Gradient Boosting model. Both algorithms were trained on standardized cross-modal features and evaluated through a strict Leave-One-Subject-Out Cross-Validation approach to guarantee predictive generalizability across participants.