Proposal for a U.S. College Basketball Player Recommendation System for the B-League

Abstract

In the B-League, Japan professional basketball league, foreign players account for a large percentage of the scoring and are the core. Therefore, the strategy of scouting talented foreign players is essential for the teams. However, team managers can not survey all players because there are many foreign leagues and a huge number of players. Therefore, most of the scouting is done through agents, so intermediary costs are significant. We propose an efficient system to find players from the NCAA (National Collegiate Athletic Association) where many of the foreign players in the B-League came from. We focus on the following topics: (1) Classify the playing styles of B-League players using k-means. (2) Estimate NCAA players’ playing styles using the B-League players’ playing style classification model. (3) Estimate efficient lineups using regression. (4) Estimate players available for scouting on a B-League budget based on scoring and age distribution. (5) Simulation for Chiba Jets Funabashi, which is a B-League team. We verified the accuracy of those using play-by-play and boxscore data of B-League from 2015-16 to 2021-22 and NCAA from 2005-06 to 2021-22 seasons. As a result, although some issues have to be resolved, we demonstrated its potential to contribute to the efficiency of player scouting.