Recent rather than enduring: opponent-related and temporal limits of situation-solution fingerprints in elite football build-up

Abstract

Version 1.0.0 provides the complete reproducibility materials supporting the study “Recent rather than enduring: opponent-related and temporal limits of situation-solution fingerprints in elite football build-up”. The archive contains derived defensive-third episode and team-match inputs, dated post-hoc protocols and amendments, event-extraction and modelling scripts, temporally held-out probabilities, team- and league-season summaries, opponent-adjustment diagnostics, observation-window curves, repeated-club cross-season transport tests, situation-level actionability analyses, figure-generation code and source tables, software requirements, reference libraries, and SHA-256 integrity hashes. The sample comprises ten European domestic league-seasons (six women’s and four men’s), 2,389 matches, 153 team-seasons, and 146,728 eligible defensive-third possession episodes. The outcome distinguishes eight first-action solution families formed by pass/carry × central/wide endpoint × forward/non-forward direction. The observable starting situation combines play origin, starting corridor, and longitudinal band; score state, time band, and venue are common adjustments. Team×situation information modestly improved forward-held-out probability estimates within a season beyond marginal team identity and population context. In these data, the frozen observation-window rule was first met at seven matches. However, the remaining team-specific gain was not supported after adjustment for opponent identity, opponent×situation, starting formations, and lagged strength; top-1 and top-2 classification did not improve; and the conditional mapping failed all three available repeated-club cross-season transport tests. The supported construct is therefore a recent, team-season-bounded conditional behavioural fingerprint. It should not be interpreted as a coach-defined tactical principle, adaptation, optimality, causal benefit, or enduring club identity. Its defensible applied use is to prioritise video review and representative scenarios while displaying opponent dependence, sample size, and uncertainty. Raw StatsBomb event JSON files are not redistributed. The archive provides derived analytical inputs, event-file integrity records, source information, and executable code. StatsBomb materials remain subject to their original terms. CC BY 4.0 applies to the authors’ original code, derived analytical materials, documentation, and figures.