Automatic event detection in association football using broadcast-derived tracking data

Abstract Player and ball tracking data derived from broadcast offers a cost-effective and scalable alternative to multi-camera optical tracking systems in football. However, the practical adoption of broadcast tracking systems depends critically on the accuracy of the data they produce. This study examined the accuracy of using broadcast-derived player and ball position data for automatic event detection for a 90 min match from the 2022 FIFA World Cup. The results were compared against events generated from a high-definition multi-camera optical tracking system (TRACAB Gen 5, ChyronHego, New York, USA) and manually tagged events from FIFA’s Data Collection Unit. The results showed that broadcast-derived auto-events, particularly from Camera 1, have the potential to reach the accuracy of multi-camera optical tracking systems for certain events. The best-case performance for most events examined in this study either exceeded, matched, or fell within 0.05 F1-score of the multi-camera system performance. However, performance varied across the systems and was limited by instances of player visibility, ball tracking errors and subjectivity in event definitions, particularly around set pieces and shots. The findings of this study highlights both the capabilities and current limitations of broadcast tracking technologies, provides guidance for their appropriate use and informs future efforts to improve data accuracy in applied contexts.