Visual Search Strategies Predict Throw Accuracy in Elite Wheelchair Curling: An Eye-Tracking and Machine Learning Approach

Visual search plays a critical role in athletic performance, yet its function in adaptive sports such as wheelchair curling remains underexplored. This study investigated how eye movement features predict throw accuracy in elite and novice wheelchair curling athletes. Thirty athletes (15 experts, 15 novices) completed standardized throw accuracy and visual search tasks, during which eye movements were recorded using the EyeLink Portable Duo system. Multiple regression and support vector machine (SVM) models were employed to analyze the relationship between gaze behavior and throw performance. Results showed that expert athletes achieved significantly higher throw accuracy and demonstrated more efficient visual search patterns, including shorter dwell times, faster reaction times, and fewer fixations. Regression analysis identified key eye movement predictors of performance, while the SVM model distinguished between expert and novice groups with 90% classification accuracy and an Area Under the Curve (AUC) of 0.93. These findings confirm a strong link between visual search efficiency and motor performance, suggesting that experienced athletes rely on optimized gaze strategies. The integration of eye-tracking and machine learning offers valuable insights for assessing performance levels and tailoring individualized training approaches in adaptive sports.