Wide Open Gazes: Quantifying Visual Exploratory Behavior in Soccer with Pose Enhanced Positional Data

A groundbreaking AI research paper titled 'Wide Open Gazes: Quantifying Visual Exploratory Behavior in Soccer with Pose Enhanced Positional Data' introduces a novel approach to measuring what players actually see on the field. Authored by Joris Bekkers, the research addresses limitations in traditional methods that rely on counting rapid head movements exceeding 125°/s, which suffer from player position bias, binary measurements, and incompatibility with existing analytics frameworks.

The methodology creates a 'continuous stochastic vision layer' using pose-enhanced spatiotemporal tracking data. By incorporating head and shoulder rotation angles, the system generates speed-dependent vision maps for individual players in a two-dimensional top-down plane. These probabilistic field-of-view and occlusion models are then combined with pitch control and pitch value surfaces to analyze both the 'awaiting phase' (when a player awaits a pass) and subsequent on-ball actions.

Using synchronized data from 32 Copa America 2024 games, the research demonstrates that aggregated visual metrics—specifically the percentage of defended area observed while awaiting a pass—are predictive of controlled pitch value gained at the end of dribbling actions. This represents a significant advancement over traditional binary 'scanning or not' measurements. The open-sourced tools enable seamless integration with existing soccer analytics frameworks, providing continuous measurements that work across all player positions without manual annotation requirements.