Reassessing the impact of shared experience in football

• Published in: Journal of big data Vol. 12; no. 1; pp. 157 - 19
• Main Authors: Le Coz, Sebastian, Carling, Christopher, Houde, Titouan, Imbach, Frank
• Format: Journal Article
• Language: English
• Published: Cham Springer International Publishing 08.07.2025; Springer Nature B.V; SpringerOpen
• Subjects: Big Data and Data-driven in Sports; Cohesion; Communications Engineering; Computational Science and Engineering; Computer Science; Confounding (Statistics); Data Mining and Knowledge Discovery; Database Management; Datasets; Decision making; Decision trees; Football; Information Storage and Retrieval; Machine learning; Mathematical Applications in Computer Science; Networks; Passes; Ratings & rankings; Shared experience; Soccer; Tournaments & championships; Cohesion; Shared experience; Football; Passes
• ISSN: 2196-1115; 2196-1115
• DOI: 10.1186/s40537-025-01239-x

This study investigates the impact of shared experience on football performance through macro-level (match outcomes) and micro-level (passing dynamics) analyses. Using machine learning models on 19,721 matches, the macro study found that conventional metrics like FIFA rankings and team strength overshadowed shared experience, which showed no independent influence on results. In contrast, micro-level analysis of 1,602 matches revealed that shared experience combined with successful prior interactions significantly predicted pass success. Decision tree models highlighted that player pairs with over 10 successful passes achieved higher completion rates, particularly for low, ground, and high passes, while pass length and receiver position further modulated outcomes. Temporal analysis showed that both short-term (20-day) and long-term (12-month) shared experience windows were predictive, with no added benefit from exponential decay weighting. The findings emphasize that shared experience’s value lies in the quality of successful coordination rather than mere familiarity or time spent together. This dual-level approach challenges assumptions about cohesion’s uniform impact, advocating for context-sensitive models that prioritize interaction quality in team performance analysis.