Analysis of players’ configurations by means of artificial neural networks

Quantifying tactics in team sports is of major interest for reliable diagnostics and goal oriented training but it is usually accompanied by the problem of considering several athletes at once. The purpose of this study is to analyze changes of patterns of configurations and the team specific variability of the underlying tactical concepts. This is done by means of self organizing maps that are trained with the configurations (i.e. players’ positions) of female volleyball players from the world championship 2002. A time discrete (target configurations) as well as time continuous (trajectories of configuration changes) oriented approach is chosen and compared. Results show that the classification of constellations is possible and may support a qualitative analysis of structural interactions within a game. Furthermore, a detailed analysis of the rally from Germany vs. Italy shows, that the German team uses fewer types of configurations more often whereas the Italian team chooses from more constellations but less frequently. Since the world champion of 2002 – Italy – shows a higher variability of configuration patterns, these variable tactics seem to correlate with success in team sports.