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

As elite women's football evolves tactically, understanding goalkeeper distribution behaviour is increasingly important for performance analysis and coaching. Traditional approaches rely on isolated passing metrics, providing limited insight into how goalkeeper actions are organised within possession structures. This study applied a graph-based clustering framework to classify goalkeeper passing behaviours in the English Women's Super League across three seasons. The dataset comprised 29,911 goalkeeper passes from 27,895 possessions. Pass embeddings were represented within a relational graph and clustered using K-means (k = 14), identifying recurrent distribution patterns shaped by spatial context, defensive pressure, and possession phase. Results may indicate that behaviours were organised along continuous gradients rather than discrete categories. Central clusters reflected adaptable distributions, whereas peripheral clusters captured more specialised actions, including long passes under defensive pressure. Lateral tendencies suggested interactions between dominant-side preferences and spatial availability. Phase analysis showed goalkeeper involvement was most prominent during transitional moments, particularly at possession initiation and termination, with comparatively limited involvement in sustained attacking sequences. Collectively, these findings provide a structured account of goalkeeper distribution in elite women's football and demonstrate how graph-based modelling can support behavioural profiling, tactical interpretation, and evidence-based coaching practice.