A Visual Analytics Approach For Soccer Tactics

Soccer is one of most popular sports over the world.Researchers have proposed different approaches to facilitate the tactical analysis of soccer and to improve team performance.The tactical analysis of soccer mainly focus on two aspects,i.e.,formations and passing,respec-tively.Formations assign players with different roles and spatial regions while passing can help a team possess the ball and seek chances for attacks.Different statistical models have been contributed to the analysis of the two important aspects.However,these models usually focus on a specific target and cannot provide a holistic view for users.More importantly,there are many unclear and unexpected data patterns within the dataset,which cannot be easily identified by statistical models.Hence,users require an interface for integrating powerful models with domain knowledge to accomplish the tactical analysis of soccer.Visualizations have been applied to soccer analysis in recent years to intuitively present the soccer data and to support interactive analysis.However,existing soccer visualizations cannot fully support the tactical analysis of soccer,particularly,the analysis of formations and passing,due to three challenges.The first challenge is about the spatio-temporal change of formations.A team could dynamically change the formation to cope with different situations The spatio-temporal nature of formations and other characteristics of soccer data,such as mul-tivariate features,make analysis of formations in soccer a challenging problem.The second challenge is about the dynamics of passing.Users hope to learn when and how would a team change their passing behaviors in a soccer match.However,existing visualizations focus on visualizing passing in a soccer phase.How to visualize the passing of hundreds of phases in a match remained to be resolved.The third challenge is about the casual analysis of soccer tactics.Most existing works adopt the correlation analysis for analyzing the relations between soccer variables.However,correlation is not causation,which motivates users to apply causal analysis to obtain more clear and confident insights.How to apply the causal analysis to soccer tactics is still an open questionIn this dissertation,we focus on studying the visual analytics approaches for soccer tactics The contributions are as follows·An approach for visualizing spatio-temporal team formations in soccer.A novel de-sign named formation flow is proposed to show the spatial-temporal change of formations and to reveal players' movements within formations.A visual analytics system,ForVizor,is further designed to help users accomplish the comprehensive formation analysis·An approach for visualizing the passing dynamics of soccer.Topic modeling is intro-duced to extract the dynamic passing patterns.Based on the patterns,this approach uses a glyph-based design to present the multi-variate information of passing of multiple phases An interactive system,PassVizor,is implemented to support the hierarchical analysis of passing·An approach for visual causal analysis of soccer.The approach provides a system,Causality Explorer,to support the examination,the validation,and the application of soccer causalities.The system use a scalable visual representation to show large scale soccer causalities and its uncertainties.Users can further achieve what-if analysis of soccer tactics with system interactions.