Goalkeeper passing behaviour in women's football: A graph-based analysis

Area of Science:

• Sports Science
• Football Analytics
• Performance Analysis

Background:

• Tactical evolution in elite women's football necessitates deeper understanding of goalkeeper distribution.
• Traditional metrics offer limited insight into organized goalkeeper actions within possession structures.

Purpose of the Study:

• To classify goalkeeper passing behaviours in the English Women's Super League using a graph-based clustering framework.
• To identify recurrent distribution patterns influenced by spatial context, defensive pressure, and possession phase.

Main Methods:

• Applied a graph-based clustering framework (K-means, k=14) to 29,911 goalkeeper passes from 27,895 possessions.
• Represented pass embeddings within a relational graph to identify behavioural patterns.

Main Results:

• Identified 14 distinct clusters of goalkeeper distribution patterns.
• Observed behaviours organized along continuous gradients, with central clusters showing adaptability and peripheral clusters specialized actions.
• Found goalkeeper involvement peaked during transitional moments (possession initiation/termination), with less involvement in sustained attacks.

Conclusions:

• Graph-based modelling provides a structured account of goalkeeper distribution in elite women's football.
• Findings support behavioural profiling, tactical interpretation, and evidence-based coaching practices.
• Results highlight the dynamic and context-dependent nature of goalkeeper distribution strategies.