Match-play data-driven player positions within the Australian Football League Women’s competition

Background/aim:
Currently, no analysis of competition technical skill data exists by player position in the Australian Football League Women’s (AFLW) competition. Understanding player positional performance roles are important for match-play tactics, player recruitment, talent identification and development. Our primary aim was, to observe what positions and roles characterise AFLW match-play using detailed technical skill action data of players. We also provide a commentary on the application of clustering methods to achieve more interpretable, reflective positional clustering results.

Methods:
A two-stage, unsupervised clustering approach was applied to match-play data from five seasons of the AFLW, totalling 1465 players. Applying clustering algorithms in an effective manner takes careful consideration with the use of principal components analysis for data reduction, and random forests for interpretation of important variables key to producing actionable results.

Results:
First-stage clustering found four positions, following common Australian Football convention of forwards, midfielders, defenders, and rucks. Key performance indicators of interest included the field location of actions across all positions, contested possessions and clearances for midfielders, interceptions and rebound 50s for defenders, hitouts for rucks, and statistics inside attacking zones for forwards. Second-stage clustering revealed 13 roles within these positions (forwards, midfielders, and rucks: three each, defenders: four). Forward clusters were nominated as a high scoring or general forward, with the third cluster representing those that also spend time in midfield. Defensive clustering had similar clusters, with high-disposal instead of high-scoring and an around-the-ground defender who performed more actions outside defence in addition to the general defender. Midfield clusters were determined by those who primarily do attacking work, defensive work, and those that perform both. Rucks clusters included general rucks, those who kick goals, and those that perform actions all around the ground.

Conclusions:
Positional roles within AFLW match-play may not be constrained to existing positional classification seen in previous men’s literature. This finding, coupled with key game actions players need to perform by position, can assist training practices, while defining new roles with suggestions of how to best use available data, producing more holistic player performance profiles. Data analysts derive benefit from the application of clustering in this environment with key data and methodological clustering and data considerations to make while producing interpretable, reproducible, comparable results assisting improved match-play performance and future research. These results will also be considered with physical performance in mind.