A complex, dynamic system called field hockey : unravelling the complexities through performance analysis

    Student thesis: Doctoral Thesis


    The performance analysis of team sports involves the process of capturing, analysing, and visualising data to communicate insights and translate these into effective strategy. Actionable insights are gained by considering game strategy from both a simple and complex perspective. However, there is often a disconnect between research and practice resulting in ineffective communication of strategy. The aim of this thesis was to develop a practical performance analysis process for field hockey, that evaluated the holistic nature of sport performance and strategy by presenting several layers of critical game information. This thesis comprises five studies: a systematic review that identified and assessed performance analysis techniques in team invasion sports; a second systematic review that assessed the use of performance analysis in field hockey; two experimental studies that captured and analysed data on game styles and ball movement patterns among international field hockey teams; and finally, the development of a web-based application to illustrate and communicate the simple-to-complex layers of performance analysis. A systematic review of the literature on all team invasion sports provided an overview of the evolution of the current best practice performance analysis techniques. Techniques included game actions, dynamic game actions, movement patterns, collective team behaviours, social network analysis, and game styles. Key factors to consider when designing a performance analysis system were game events in relation to time, space, opposition, and match context. However, these factors must be presented in simplified, practical profiles by analysing their interactions. The second systematic review, focused solely on performance analysis techniques and considerations in field hockey, was completed to identify gaps in the literature and the suitability of performance analysis techniques used in other sports to be employed in field hockey. Techniques identified were limited to outcome-focused approaches which reflects restricted resources and time available afforded to performance analysts in field hockey. The analytic method considered most effective to employ in field hockey based on the findings from Study 1 and Study 2 was game styles analysis. Game styles identify the consistent strategy implemented by a team through the clustering of similar dynamic game actions. This technique was deemed the most appropriate as game data are captured using basic equipment, and clustering algorithms eliminate the time needed for an analyst to detect patterns in the data manually. The next two experimental studies were undertaken to quantify game styles using different perspectives. Game styles were identified by capturing and analysing in-game events, and ball movement patterns. Video footage of 131 games (n = 74 women, n = 57 men) from the 2019 Pro League international field hockey tournament were reviewed using a computerised notational analysis system. In-game events captured included game actions, stoppages, turnovers, and goal shots in relation to time, space, and the opposition. Variables were normalised by converting to percentages or ratios and divided into six categories of performance. A k-means cluster analysis was performed on each category, and two clusters were identified for each, reflecting the opposing game style types that could be used. Game style types included strong or poor for established and counter attack success, high or low for set piece occurrence, dribbling or passing for established and counter attack game actions, and direct or possession for tempo. The percentage of games each team implemented a particular game style was calculated, and a profile developed by identifying the game style types used consistently by a team. The influence of contextual factors on game style types was assessed using decision trees with opposition game style type, and other reference team game style categories having the greatest effect on performance. For example, for women, a team was 4 times more likely to be poor in established attack if the opposition were strong in counter attack; nearly twice as likely to pass in established attacks if they dribbled in counter attacks; and twice as likely to play direct if they were poor in counter attacks. The second experimental study analysed ball movement patterns to identify how and where a team moved the ball to create goal scoring opportunities. A field was divided into 40 cells, and the start and end locations of ball movements were recorded, as well as play outcomes. Raw x, y data were converted to possession by calculating time for ball movements in each cell, entropy (a measure of variation/unpredictability in ball movements) was calculated by determining the likelihood of the ball moving from one cell to another, and progression rates calculated by considering the direction of ball movement between cells. The 40 cells were grouped into 7 attacking zones from deep in defence, to building an attack and creating a goal scoring opportunity. Collectively, these measures were used to provide practical insight into the strategies occurring in each phase of play. The influence of match context on movement patterns was assessed using linear mixed models, however few substantial effects were observed. Decision trees identified the most important variables determining play outcomes were possession in the circle (Goal Shot: 12.6% vs. Turnover: 1.2%), movement direct to goal from deep attack (Goal Shot: 41.8% vs. Turnover: 11%), and entropy in build attack (Goal Shot: 0.12 vs Turnover: 0.44) and build defence (Goal Shot: 0.10 vs. Turnover: 0.50). A direct approach through the centre of the field was more likely to lead to a goal shot. However, limited goal scoring opportunities within a game indicates teams should be unpredictable to maintain possession until the opposition become unbalanced opening the opportunity to attack. The trivial effect of match context on game styles and ball movement patterns reflects that there is more than one way to win a game, and teams need to develop strategies based on their strengths and ability to exploit the oppositions’ weaknesses. These outcomes led to the final study; the development of a web-based application to visualise and communicate the analysis from the two experimental studies to showcase the similarities and differences, and strengths and weaknesses of individual teams. A Shiny application was developed in RStudio using the shiny R package, which produces an interactive web platform that is hosted and shared online. Each discrete step in the performance analysis process underpinned the development of the application. A central element is visualisations from holistic game style and ball movement profiles to technical-tactical indicators reflecting specific passages of play. The performance analysis process used in essence provides complexity to simple variables while simplifying the complex variables. The Shiny application forms a useful holistic and layered approach to understanding, developing and communicating strategy in field hockey and is freely accessible to performance analysts and coaches to implement findings in practice. This thesis has assessed the practicality of performance analysis techniques in team invasion sports, and developed an evidence-based approach that can be used in field hockey based on readily-available resources. Novel approaches were utilised to develop game style and ball movement profiles to understand strategy by considering the effect of the opposition, time, and space. Visualisations accessed through a web-based application were used to connect the simple and complex variables surrounding strategy. The performance analysis process developed allows the complexity of strategy to be unravelled, facilitating performance analysts and coaches to better understand, develop and communicate strategy in practice.
    Date of Award2022
    Original languageEnglish
    Awarding Institution
    • University of Canberra
    SupervisorJocelyn Mara (Supervisor), David Pyne (Supervisor) & Marijke Welvaert (Supervisor)

    Cite this