With the steady reduction in the price of storage, and increasing availability of high quality recording devices, much effort has been invested in investigating methods to index large collections of high dimensional datasets. Archives of sporting events are well represented within this set of large datasets. Most efforts to index sport related data have concentrated on the indexing of collections of audio/video data. This thesis presents and evaluates several novel methods to index football matches based on the underlying trajectory of the ball and players, rather than the raw video. This allows for the potential of very expressive indexing systems. The second strand of this thesis explores the use of the underlying trajectory data to build behavioural models of players. A promising hierarchical approach is undertaken, whereby the behaviour of individual players is influenced by the cliques of players they associate with, as well as the team as a whole. Although both the indexing and behavioural modelling aspects of this thesis use data from football as the basis for the work, in principle the approaches taken are general enough to apply to any team based game.