Complex networks untangle competitive advantage in Australian football

Research output: Contribution to journalArticle

1 Citation (Scopus)
111 Downloads (Pure)

Abstract

We construct player-based complex network models of Australian football teams for the 2014 Australian Football League season; modelling the passes between players as weighted, directed edges. We show that analysis of these measures can give an insight into the underlying structure and strategy of Australian football teams, quantitatively distinguishing different playing styles. The relationships observed between network properties and match outcomes suggest that successful teams exhibit well-connected passing networks with the passes distributed between all 22 players as evenly as possible. Linear regression models of team scores and match margins show significant improvements in R2 and Bayesian information criterion when network measures are added to models that use conventional measures, demonstrating that network analysis measures contain useful, extra information. Several measures, particularly the mean betweenness centrality, are shown to be useful in predicting the outcomes of future matches, suggesting they measure some aspect of the intrinsic strength of teams. In addition, several local centrality measures are shown to be useful in analysing individual players' differing contributions to the team's structure.

Original languageEnglish
Article number053105
JournalChaos
Volume28
Issue number5
DOIs
Publication statusPublished - 1 May 2018

Fingerprint

Complex networks
Complex Networks
Electric network analysis
Linear regression
Centrality
network analysis
Bayesian Information Criterion
Betweenness
Network Analysis
Linear Regression Model
Margin
Network Model
regression analysis
margins
Modeling

Cite this

@article{8ec52a506e2246019f2c09e52f992981,
title = "Complex networks untangle competitive advantage in Australian football",
abstract = "We construct player-based complex network models of Australian football teams for the 2014 Australian Football League season; modelling the passes between players as weighted, directed edges. We show that analysis of these measures can give an insight into the underlying structure and strategy of Australian football teams, quantitatively distinguishing different playing styles. The relationships observed between network properties and match outcomes suggest that successful teams exhibit well-connected passing networks with the passes distributed between all 22 players as evenly as possible. Linear regression models of team scores and match margins show significant improvements in R2 and Bayesian information criterion when network measures are added to models that use conventional measures, demonstrating that network analysis measures contain useful, extra information. Several measures, particularly the mean betweenness centrality, are shown to be useful in predicting the outcomes of future matches, suggesting they measure some aspect of the intrinsic strength of teams. In addition, several local centrality measures are shown to be useful in analysing individual players' differing contributions to the team's structure.",
author = "Calum Braham and Michael Small",
year = "2018",
month = "5",
day = "1",
doi = "10.1063/1.5006986",
language = "English",
volume = "28",
journal = "Chaos",
issn = "1054-1500",
publisher = "ACOUSTICAL SOC AMER AMER INST PHYSICS",
number = "5",

}

Complex networks untangle competitive advantage in Australian football. / Braham, Calum; Small, Michael.

In: Chaos, Vol. 28, No. 5, 053105, 01.05.2018.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Complex networks untangle competitive advantage in Australian football

AU - Braham, Calum

AU - Small, Michael

PY - 2018/5/1

Y1 - 2018/5/1

N2 - We construct player-based complex network models of Australian football teams for the 2014 Australian Football League season; modelling the passes between players as weighted, directed edges. We show that analysis of these measures can give an insight into the underlying structure and strategy of Australian football teams, quantitatively distinguishing different playing styles. The relationships observed between network properties and match outcomes suggest that successful teams exhibit well-connected passing networks with the passes distributed between all 22 players as evenly as possible. Linear regression models of team scores and match margins show significant improvements in R2 and Bayesian information criterion when network measures are added to models that use conventional measures, demonstrating that network analysis measures contain useful, extra information. Several measures, particularly the mean betweenness centrality, are shown to be useful in predicting the outcomes of future matches, suggesting they measure some aspect of the intrinsic strength of teams. In addition, several local centrality measures are shown to be useful in analysing individual players' differing contributions to the team's structure.

AB - We construct player-based complex network models of Australian football teams for the 2014 Australian Football League season; modelling the passes between players as weighted, directed edges. We show that analysis of these measures can give an insight into the underlying structure and strategy of Australian football teams, quantitatively distinguishing different playing styles. The relationships observed between network properties and match outcomes suggest that successful teams exhibit well-connected passing networks with the passes distributed between all 22 players as evenly as possible. Linear regression models of team scores and match margins show significant improvements in R2 and Bayesian information criterion when network measures are added to models that use conventional measures, demonstrating that network analysis measures contain useful, extra information. Several measures, particularly the mean betweenness centrality, are shown to be useful in predicting the outcomes of future matches, suggesting they measure some aspect of the intrinsic strength of teams. In addition, several local centrality measures are shown to be useful in analysing individual players' differing contributions to the team's structure.

UR - http://www.scopus.com/inward/record.url?scp=85046900252&partnerID=8YFLogxK

U2 - 10.1063/1.5006986

DO - 10.1063/1.5006986

M3 - Article

VL - 28

JO - Chaos

JF - Chaos

SN - 1054-1500

IS - 5

M1 - 053105

ER -