[2D1-OS-6] 移動系列のデータマイニングと機械学習(1/1)

[2D1-OS-6-05] Evaluation of group behavior sequences based on a prediction model: Application to soccer team defense

Kosuke Toda², Masakiyo Teranishi¹, Keisuke Kushiro², 〇Keisuke Fujii^1,3,4 (1. Nagoya University, 2. Kyoto University, 3. RIKEN, 4. JST Presto)

Keywords:Sports, Moving sequences, Multi-agent

With the development of measurement technology, data on the movements of actual games in various sports has become available and is expected to be used for strategy and evaluation. In particular, defenses in team sports are generally difficult to be evaluated because they are played as a team and their statistics are not often recorded. Conventional evaluation methods based on predictions of scores are considered to be unstable because they predict rare events in the entire game, and it is difficult to evaluate various plays leading up to the score. On the other hand, evaluation methods based on certain plays that lead to scoring and dominant region are sometimes difficult to evaluate players and teams in relation to their overall performance (e.g., points scored). In this study, we propose a method for evaluating a team's defense from a comprehensive perspective related to the team's performance, based on the prediction of ball recovery and being attacked, which occur more frequently than goals, using player behavior and positional information of all players and the ball. Using data from 45 soccer matches, we examined the relationship between our index and the team's performances in actual matches and the season.