Analysing strategy and tactics is an integral part of modern basketball performance analysis. Until recently however, the majority of research focused purely on outcomes rather than the behaviour that led to these outcomes. More specifically, the majority of research to date has focused on analysing counts of performance indicators, such as assists, rebounds, freethrows, made baskets etc, and while this provides some value, they tell us little about how those indicators were generated and fail to provide coaches with actionable insights. Consequently, the aim of this thesis was to investigate coach-driven questions focused on developing performance intelligence by analysing team strategy and tactics, otherwise known as `collective behaviour`, in international women`s basketball. The current thesis is presented as individual chapters containing stand-alone studies which have been or will be submitted to peer-reviewed academic journals for publication. Following a general introduction in chapter 1, a review of literature is presented in the second chapter detailing all relevant research that had attempted to measure or analyse tactics and/or strategy in any team sport prior to the first study of the current thesis (August 2015). The review identified two distinct types of research; (i) studies that focused on developing new methods for analysing strategy and tactics, and (ii) research focused on presenting tactical findings, often using previously developed methods. In addition, due to technological advancements, studies differed significantly in terms of the data utilised. With the advent and application of spatio-temporal tracking, a shift was emerging from observational data to spatio-temporal data. This led to a range of new methodologies and applications for performance analysis. With the increased utilisation of spatio-temporal tracking data, there was a shift in focus of research to one that sought to understand collective behaviour in team sport, from analysing team spatial characteristics and constellations, to mapping the most successful team movement patterns, and attempting to fit player movements into a dynamic systems framework. The findings of the literature review were used to direct the series of studies that follow in this thesis, which are comprised of a mix of both methodological and practical investigations using spatio-temporal ball tracking data. Furthermore, to maximise practical implications and impact, we focused on answering coach-driven questions through a partnership with Basketball Australia and the Australian Institute of Sport. This resulted in the eventual theme of the thesis, examining unpredictability in basketball and its effects on performance. Consequently, in chapter three we present a method for measuring spatial entropy across the basketball court and apply the method to ball-tracking data collected for six international women`s basketball teams. The study determined that winning teams played with significantly higher entropy in the front-court than losing teams. Having defined a spatial measure of unpredictability that was associated with outcome measures of team performance, chapter four established a spatial measure of performance for which we could attempt to test the relationship between the two variables. In addition, a novel measure of performance was developed utilising tracking data capable of quantifying spatial effectiveness of ball movement as well as identifying common ball movement sequences. This study identified which areas of the court were the most effective for each team, irrespective of where or when the shot was taken. This highlighted the strengths and weaknesses of different teams, and provided insight into specific ball movement patterns and their distributions for each team. The results provide evidence of team specific strategies for scoring, with team specific ball movement sequences further adding to available performance intelligence. Chapter five presents an appropriate statistical method for comparing the previously established spatial measures of performance. A Bayesian hierarchical model was introduced. These models are common in other fields, but as yet, to our knowledge, had not been applied to team sport research. These models are ideal for data with spatial dependency - meaning data in which the regular statistical assumptions of independence amongst observations are violated due to spatial nature of the data. Furthermore, we demonstrated how different spatial scales can significantly impact the results, and therefore, the conclusions drawn from these results. We tested a 6x6, 4x4, and basketball-specific spatial scale. The results from these analyses showed a generally positive association such that increases in spatial entropy were associated with increases in spatial effectiveness. That is, teams that played with less predictability were more successful. Furthermore, we found the strength and confidence of the associations between the two variables were impacted by the spatial scale used, with the 6x6 and 4x4 grids showing the strong positive relationships, while the basketball-specific scale was less conclusive. In chapter six, with evidence of an association between our measures of entropy and effectiveness, we further refine these measures to compare the final five seconds of ball movement entropy to shooting efficiency. This study provided a more focused investigation of the impact of entropy on performance by considering only the final sequence of ball movements leading to a shot. We hypothesised that greater unpredictability of ball movements prior to a shot would lead to more positive outcomes. The results indicated each team performed differently, with Australia, Japan, France, and the USA showing definitive positive relationships between entropy and scoring efficiency, while Belarus and Turkey showed less conclusive evidence of this relationship. This finding may be due to differing styles of play between teams. In the discussion chapter (Chapter 7) we review the findings of each study, discuss the new knowledge that has emerged from the thesis, outline their practical implications, discuss future directions for research on this topic, and finally, provide concluding remarks for the thesis.
© Copyright 2019 Veröffentlicht von University of Sydney. Alle Rechte vorbehalten.