Abstract
Visual search is pivotal for athletic performance, yet its role in adaptive sports like wheelchair curling remains understudied. This study investigated how eye-movement features predict delivery accuracy and distinguish elite from novice athletes. Thirty wheelchair curling athletes (15 experts, 15 novices) performed standardized delivery accuracy and visual search tasks, with eye movements recorded using the EyeLink Portable Duo system. We employed multiple regression to identify predictors of accuracy and a support vector machine (SVM) to classify athletes based on expertise. Experts demonstrated superior delivery accuracy and significantly more efficient visual search patterns, characterized by shorter dwell times, faster reaction times, and fewer fixations. The SVM model successfully classified athletes with 90% accuracy (AUC = 0.93), while regression analysis confirmed that specific gaze metrics were robust factors associated with performance. These findings establish a strong quantitative link between efficient gaze strategies and expert motor performance in a constrained-mobility setting. This integrated eye-tracking and machine learning approach offers a powerful framework for objectively evaluating performance and developing data-driven, personalized training interventions in wheelchair curling and other precision-focused adaptive sports.