Achievement Standards and Causal Structure of Offensive Skill Measurement Items in Japan Professional Football League

抄録

Since it is not easy for analysts in the instruction field to process the performance big data of soccer games, it is necessary to develop models and analysis methods using big data and return them to the instruction field. Game performance data is measured by a quantitative scale, but if it is converted into achievement type binary data, the measurement cost is reduced, and a criterion-referenced evaluation scale with IRT applied can be constructed. However, converting to binary data reduces the amount of information, so it is necessary to reconfirm whether it reflects soccer skills. Therefore, this study’s purpose was to clarify the causal structure of game performance big data with binary data in soccer. To that end, we analyzed the achievement standards for constructing binary data using quantitative game performance big data in J.League offensive plays. The achievement standards were calculated by CART using the Gini impurity, and achievement data converted into a binary scale was constructed. As a result of applying the maximum likelihood method and factor analysis of oblimin rotation to the achievement data and clarifying the factor structure, seven subfactors of offensive skill were extracted. As a result of a path analysis from the variance-covariance matrix of the factor scores estimated by the maximum likelihood method, the causal structure that reflects the attacking style of soccer was clarified in the achievement data. The factor structure and causal structure were superior to the quantitative data.