In this paper, we propose a method which can enable us to visualize relationships between shoot and formation components for team sports. In the proposed method, a dictionary learning for nonnegative sparse coding is preformed to obtain a dictionary of base player densities, which can be considered as components of player formations. Then a nonnegative sparse coding evaluates coefficient vectors as features in corresponding scenes. After that, a co-occurrence analysis based on mutual self-information draws a co-occurrence network for shoot and formation components. For validation, the proposed method is applied to player's position data of five matches in 2011 J. League.
