Predictive frameworks for youth basketball: a leakage-aware reassessment of athlete development and injury-risk models
- 1Nanjing Institute of Technology (NJIT), Nanjing, China
Abstract
Introduction: Artificial intelligence is increasingly applied in sports analytics; however, exceptionally high predictive performance may result from target construction, data leakage, or deterministic label structures. This study reassessed predictive models for youth basketball using a leakage-aware framework.
Methods: A publicly available youth-basketball dataset containing 1,000 records and 25 variables was analyzed. Development-score construction and injury-risk label structure were audited. Construction-related predictors were subsequently excluded, and model performance was evaluated using repeated cross-validation and a locked 200-athlete holdout. Subgroup analyses by gender and age and an external-validation feasibility assessment were also conducted.
Results: Development_Score was reconstructed almost perfectly from its documented components, with repeated cross-validation R2 of 0.999925. Injury_Risk_Level also demonstrated deterministic structure based on strength and endurance. After excluding construction-related predictors, locked-holdout performance decreased to 0.750 accuracy, 0.329 balanced accuracy, 0.286 macro-F1, and 0.622 macro AUC. High-risk sensitivity was 0.000 across the overall sample and examined subgroups. No compatible public dataset was identified for external validation.
Discussion: The initially exceptional performance largely reflected target construction and deterministic labeling rather than independent predictive information. These findings emphasize leakage-aware validation, transparent reporting, and cautious interpretation of sports-analytics prediction models.
Related
- Brief Report: Reassessment of bTMB as a Predictive Biomarker in NSCLC Immunotherapy—A Reanalysis of OAK and POPLAR
- Review of human health exposure assessment frameworks in the United States and development of a unified framework for exposure assessment practitioners
- Development and validation of a nomogram for predicting anastomotic leakage after esophagectomy: integrating baseline characteristics and early postoperative biomarkers.
- Risk factors for plastic bronchitis complicating severe Mycoplasma pneumoniae pneumonia in children and development of a machine learning-based predictive model
- Quality by Design (QbD) Empowered by IVIVR Modelling: A Predictive Approach for Drug Formulation Development
BibTeX
@article{yao2026predictive,
title = {{Predictive frameworks for youth basketball: a leakage-aware reassessment of athlete development and injury-risk models}},
author = {Chen Yao},
journal = {Frontiers in Physiology},
year = {2026},
month = oct,
doi = {10.3389/fphys.2026.1912035},
url = {https://doi.org/10.3389/fphys.2026.1912035},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Predictive frameworks for youth basketball: a leakage-aware reassessment of athlete development and injury-risk models
AU - Chen Yao
PY - 2026
DA - 2026/10/22
JO - Frontiers in Physiology
DO - 10.3389/fphys.2026.1912035
UR - https://doi.org/10.3389/fphys.2026.1912035
AB - Introduction: Artificial intelligence is increasingly applied in sports analytics; however, exceptionally high predictive performance may result from target construction, data leakage, or deterministic label structures. This study reassessed predictive models for youth basketball using a leakage-aware framework. Methods: A publicly available youth-basketball dataset containing 1,000 records and 25 variables was analyzed. Development-score construction and injury-risk label structure were audited. Construction-related predictors were subsequently excluded, and model performance was evaluated using repeated cross-validation and a locked 200-athlete holdout. Subgroup analyses by gender and age and an external-validation feasibility assessment were also conducted. Results: Development_Score was reconstructed almost perfectly from its documented components, with repeated cross-validation R2 of 0.999925. Injury_Risk_Level also demonstrated deterministic structure based on strength and endurance. After excluding construction-related predictors, locked-holdout performance decreased to 0.750 accuracy, 0.329 balanced accuracy, 0.286 macro-F1, and 0.622 macro AUC. High-risk sensitivity was 0.000 across the overall sample and examined subgroups. No compatible public dataset was identified for external validation. Discussion: The initially exceptional performance largely reflected target construction and deterministic labeling rather than independent predictive information. These findings emphasize leakage-aware validation, transparent reporting, and cautious interpretation of sports-analytics prediction models.
SN - 1664-042X
ER -