Determining spatio-temporal metrics that distinguish play outcomes in field hockey

Tactical behaviour in field sports can be examined using spatio-temporal metrics, which are descriptions of player behaviour derived from data of player positions over time. Many metrics can be computed that describe the cooperative and adversarial interactions between players. The methods typically used by sports performance analysts cannot appropriately analyse the many possible spatio-temporal metrics and their interactions. Tantalisingly, the interactions between these descriptions of player behaviour could potentially describe tactical differences in performance. This thesis describes a programme of research that determined some spatiotemporal metrics that distinguish play outcomes in field hockey. Methods inspired by genetic analysts were used to estimate the influence of combinations of spatio-temporal metrics on the outcome of field hockey plays. The novel application of the genetic methods to sports performance data raised some practical difficulties. Adjustments to the method facilitated the selection of distinguishing metric combinations from an initially large list of over 3,600 metrics. The adjustments made to the genetic methods represent one of several contributions to knowledge made by this programme of research. These contributions will help performance analysts with the increasingly common task of analysing high-dimensional data. Other contributions to knowledge are a suite of metric combinations that distinguish play outcomes in field hockey and empirical support for some tactical preconceptions. The key finding of interest for players and coaches is that play outcomes in field hockey are distinguished by proximity to the goal and passing execution. The metrics that distinguish the several outcomes differ depending on the outcomes being compared. Coaches and athletes should therefore recognise the variety of tactics required to minimise negative outcomes and maximise positive ones.