I already mentioned that issue of IEE Pervasive Computing journal about sports. I just read the on about football: Computerized Real-Time Analysis of Football Games by Michael Beetz, Bernhard Kirchlechner and Martin Lames IEEE Pervasive Computing, pp. 33-39. The authors present Football Interaction and Process Model (which uses a real-time positioning system), an application that acquires action models, infer action-selection criteria, and identify player and team strengths and weaknesses.

An example of how it works:

The FIPM game analyzer uses player positions to infer player roles: (a) a team’s tactical lineup; (b) motion distances (visualized as the size of the circles) and motion profiles of the players; (c) prototypical preparation of shots. The sequences of ball actions are spatially clustered.

Another example:

Analyses of ball actions. (a) Players’ action profiles. The pie charts show the fractions of passes, dribbles, and shots. (b) Passers and receivers and the number of passes between two players. The number of passes from one player to another is depicted by the width of the corresponding arrow. (c) Spatial distribution of the passes three players made.