Multivariate point processes for modelling association football matches

We consider the modelling of sequences of multivariate point processes where the occurrence rate depends on past occurrences within the process. Building on a traditional model for point processes, the Hawkes process, the main idea is to take advantage of the decomposition that motivated partial likelihood to separate the modelling of the event types and the occurrence times. We present an application on the modelling of event-sequences in football, where match events can be treated as a multivariate spatio-temporal point process. The aim is to provide inferences about previously unquantified measures governing the dynamics of the game as well as predicting the occurrence of events of interest, such as goals, in a specified interval of time.