Abstract
Recently, multiagent systems has attracted much attention. RoboCup Soccer Simulation is treated as a test bed of these. In real soccer, the analysis of games in terms of team behavior has been studied, however, it has not been done in RoboCup Soccer Simulation. In this study, we aim to facilitate the analysis of the team behavior and to clarify the involvement of the team possession type in result of games in RoboCup Soccer Simulation. In particular, we construct a method of detecting counter attack, one of team possession type and we analyze the log files of games. After considering the feature quantities to detect, we detect counter attack using support vector machine based on four feature quantities. As a result, this detector can detect in 77%. Using this detector will reduce the burden of check the log data in visual.
