Abstract
Finding and acquiring promising talent has always been one of the most important management issues in various organizations. In professional sports, such an activity is called scouting, but for a long time, this activity has been left to the intuition and experience of individuals. In order to make scouting more rational, it is important to first quantitatively evaluate the skills of potential players. Traditional methods for evaluating player ability have focused on individual-level estimation, but an approach that evaluates ability in terms of compatibility with other players is possible. In this study, we propose a scouting framework based on Factorization Machines, used in recommendation systems and known for the ability to excel in the interactions between elements for ability estimations that take into account the compatibility between players. Experimental results on professional basketball leagues show that the proposed method is more realistic than existing methods and can be effectively used for player scouting. Widespread use of the system based on the proposed method is expected to improve the efficiency of scouting, increase the liquidity of the player market, and reduce the mismatch between teams and players, thereby increasing the level of competition and revitalizing the professional sports industry.
