Collective behaviour monitoring in football using spatial temporal and network analysis: application and evaluations.

Corsie, Martin Richard

Abstract

Analysis is an important part of understanding and exploiting performance of football teams. Traditional approaches of analysis have centred around events that may not fully incorporate the highly dynamic nature of matches. To circumvent this weakness, applications of collective behaviour metrics applying spatial temporal and social network analyses to data in football have been trending over the last 10 years. The aims of this PhD were to: 1) establish the strengths and limitations of current research investigating collective behaviour in football applying novel analytical procedures; 2) investigate the credibility of present methods informing coaching practice; and 3) provide guidance for practitioners in implementing complex analytical procedures with current data collection methods. These aims were achieved through the completion of five interlinked studies. The first two studies comprised systematic reviews evaluating the quality of previous research investigating collective behaviours. The first systematic review focussed on spatial temporal metrics and the second systematic review focussed on social network analysis metrics. In addition to standard review procedures, both systematic reviews included analyses of author quotes regarding the metrics used within each study. These included description and conceptualisation of each metric, along with practical applications and measurements of reliability. The first systematic review identified several limitations in the current literature base of spatial temporal metrics investigating collective behaviour in football. These included a lack of conceptualisation of the metrics used, assumptions of metric reliability, frequent use of broad and non-actionable practical recommendations, failure to justify sample sizes and a bias towards including males. Similar findings were found in the social network analysis systematic review where authors also seldom conceptualised metrics, provided vague practical applications and often failed to justify sample size. Literature including social network analysis were also inconsistent with the metric calculations and nearly all studies investigated elite male matches. The third study in this PhD attempted to quantify the reliability of spatial temporal metrics by simulating expected error values on top of real-world data. Through fitting linear mixed effects models on signal to noise ratios, metrics were established to be reliable where positioning systems are accurate to 0.5 m or less. In situations where positioning systems errors were approached 2 m, only some were considered to produce reliable values, (e.g. team centroid), whereas metrics using distances and numerical relations were considered to produce unreliable values. After assessing the literature and reliability, the PhD focussed on implementation of common and reliable metrics, leading into the final study of the PhD which employed an iterative design comprising multiple interviews to investigate coach perceptions of collective behaviour metrics. A thematic analysis identified themes that closely resembled the 10 traditional principles of play in football, further establishing their validity. Moreover, coaches reacted positively to presented measurements, most notable network intensity, distance between defenders, triads, team length, and team depth. Coaches stated they trained players with the concepts these measurements represent as a central focus. The PhD work was concluded with a final chapter set as pedagogical support for practitioners wishing to implement these techniques providing a guide to measuring the tactical concepts discussed within this thesis. Collectively, this PhD highlights that novel collective behaviour metrics have a place in current performance analysis systems in football. Additionally, a methodology is presented for practitioners to apply to their own teams and generate specific metrics relevant to the teams own tactical principles.

Citation

CORSIE, M.R. 2022. Collective behaviour monitoring in football using spatial temporal and network analysis: application and evaluations. Robert Gordon University, PhD thesis. Hosted on OpenAIR [online]. Available from: https://doi.org/10.48526/rgu-wt-1880207

• Thesis Type: Thesis
• Deposit Date: Feb 10, 2023
• Publicly Available Date: Feb 10, 2023
• DOI: https://doi.org/10.48526/rgu-wt-1880207
• Keywords: Sport performance analysis; Athletic performance analysis; Football players; Soccer players; Network analysis; Data visualisation; Dynamic systems theory
• Public URL: https://rgu-repository.worktribe.com/output/1880207
• Award Date: Sep 30, 2022