From a fine margin to a firm advantage.

The incorporation of our event data in our computer vision model enhances player identification by adding ball possession information on top of jersey number recognition. This approach allows for a more detailed portfolio of player characteristics that is used by the algorithm to identify each player. Upon integrating event data, recognition rates are up to 98% of players with 92% precision, with a mean absolute error of 0.66 meters in player locations compared to optical tracking data. These results underscore the substantial improvement in tracking data quality achieved through our unique solution incorporating event data.