A Clustering-Based Athlete Recommender System Considering the Discrepancy Between Ability and Result

In recent years, data analysis in the sports industry has significantly advanced, enabling the analysis of players from various perspectives. The introduction of systems utilizing cutting-edge technology has allowed the quantification of player abilities, revealing players who possess high abilities but whose performance is stagnating according to the data. In this study, we hypothesized that such players, when viewed from a long-term perspective, will see their performance converge to their abilities. Based on this hypothesis, we propose a method to identify players who exhibit a discrepancy between their abilities and performance by separately clustering them based on ability and performance. Despite belonging to clusters with high abilities, these players are found to belong to clusters with low performance. We conducted experiments on MLB outfielders, and the results revealed that many of the identified players showed growth in the following year, confirming the presence of players with high potential for the future.