Outlier detection of skill decision differences of football penalty takers from a data-driven perspective: multialgorit hm modeling and empirical research of eye-tracking data.

Defensive performance in football penalty kicks relies heavily on a goalkeeper's visual anticipation and rapid decision-making capabilities. While identifying the visual search advantages of expert athletes is crucial for optimizing training, conventional eye-tracking metrics often fail to capture the complex spatial distribution patterns of gaze points. Integrating machine learning anomaly detection algorithms offers a novel methodological perspective for the precise quantification and analysis of visual strategy variations across different expertise levels. Objective: This study aimed to investigate the differences in visual search characteristics between expert and amateur football players during penalty defense tasks, leveraging eye-tracking technology and multiple machine learning algorithms. Methods: Gaze data were collected from 10 professional/semi-professional athletes (Expert Group) and 10 collegiate athletes (Amateur Group), resulting in 120 valid records. The Shapiro-Wilk test indicated that the gaze point coordinates ( Results: (1) DBSCAN results revealed a highly significant difference in outlier rates between the two groups Conclusion: Expert athletes exhibit more concentrated and stable visual search strategies (characterized by lower gaze outlier rates) during penalty defense compared to amateurs. The DBSCAN algorithm and composite indicators effectively quantify expertise-related visual search features, providing empirical evidence for understanding cognitive patterns and enhancing scientific training for football goalkeepers.