INTRODUCTION:
While the physical demands of elite women's football have been extensively quantified, the causal pathways linking running performance to match success remain elusive. Previous studies utilizing FIFA's confidential tracking data (Bradley, 2024) benchmarked positional demands but failed to elucidate structural relationships between physical, tactical, and technical variables. Conversely, recent clustering analyses using FIFA Match Summary Reports (Shen et al., 2026) classified player profiles but ignored interdependencies among indicators. To bridge these gaps, this study pioneers causal discovery algorithms to trace pathways from physical exertion to victory using exclusively open-access data from the 2023 FIFA Women's World Cup.
METHODS:
All 64 match summary reports were processed using OCR and NLP pipelines, with extracted variables stored in hierarchical JSON format to ensure data integrity and reproducibility. The dataset comprised six physical metrics (total distance and five speed zones), 24 team shape variables, and three technical outcomes (possession, goals, match result). Following descriptive and correlation analyses, we employed the PC algorithm to construct a causal graph. Bootstrap resampling (n=1000) validated edge stability and directionality.
RESULTS:
The dataset contained 22,037 leaf key-value pairs from 619 players across 32 teams, substantially exceeding the scale of prior studies relying on manual spreadsheet transcription. The causal graph revealed a tripartite structure connecting physical, formation, and result variables. Only Zone 2 (7-13 km/h) and Zone 5 (>=23 km/h) distances showed direct adjacency to team shape variables. The analysis uncovered a counter-intuitive feedback loop where match outcome exerted reverse causal influence on ball possession. Bootstrap validation demonstrated high edge stability (>=92% frequency) for key pathways including Total Distance -> Zone 2 Distance -> Defensive Low Depth -> Possession.
CONCLUSION:
Notably, Zone 2 running emerged as a critical mediator between physical load and tactical formation—a relationship previously unreported in women's football literature. This low-to-moderate intensity activity appears to serve as a tactical bridge, distinguishing our causal approach from clustering methods that treat variables as isolated features. This work makes two primary contributions: first, expanding elite women's football data resources through automated JSON dataset construction from public reports; second, demonstrating that causal discovery reveals actionable pathways where statistical and machine learning methods fail. These findings provide evidence-based insights for physical preparation and tactical planning, offering coaches structural understanding of how running performance translates into match success.