Prioritizing next-day readiness decline in elite female soccer using readiness, workload, and wellness data: A temporally validated and calibrated monitoring framework.

Abstract

Daily athlete monitoring is widely used in elite soccer, but its value depends on whether routine data can help identify athlete-days more likely to be followed by self-reported next-day readiness decline. This retrospective study evaluated a temporally validated and calibrated model for prioritizing staff-review attention in elite female soccer. Using SoccerMon data, we analyzed 14,428 eligible athlete-days from 50 players. The outcome was same-athlete next-calendar-day self-reported readiness decline ≥0.5 athlete-specific SD calculated from at least five observed readiness values strictly before the index date. Models were trained in 2020 and evaluated in a held-out 2021 test set. Adding workload and wellness to readiness improved discrimination (Model 4 vs Model 0: AUC 0.7645 vs 0.7043; PR-AUC 0.5135 vs 0.3808); paired athlete-level bootstrap ΔAUC was 0.0592 (95% CI 0.0288-0.0891). Platt recalibration reduced Brier score from 0.1924 to 0.1460 without changing ranking. Reviewing the top 20% (1,672 athlete-days) captured 881 events (46.1%) versus 36.4% for Model 0. These retrospective results support cautious staff-review prioritization rather than demonstrated improvement in athlete management or performance, and prospective implementation testing and independent external validation are required before real-world use.