srjmdSRJMDSRJMD2788-9467https://doi.org/10.47310/srjmd.2025.v05i02.003Machine Learning-Driven Selection of Optimal Physical Fitness Tests for Talent Identification in VolleyballAmerMishaal FaihanAffiliation Republic of Iraq Ministry of Education General Directorate of Education in Anbar, IraqTalent identification is a central focus of performance development in volleyball but the current methods used are not always well justified and take a long time to administer. The paper will suggest a machine-learning-based framework that would help to optimize the choice of physical fitness tests to find potentially good volleyball players. Sample size used in the study was comprised of around 200 athletes of various ages, involving anthropometric, physiological and skill-related parameters. A complete test battery, i.e., vertical jump, agility T-test, handgrip strength, VO 2 max and spike speed was applied. Random Forest, Support Vector Machine (SVM) and XGBoost machine learning models were trained to reveal the most predictive variables related to a high-level of volleyball game performance. To guarantee robustness, cross-validation was used and the performance of the model was measured using ROC and AUC values. The findings indicated that a subset of physical fitness tests is able to produce similar predictive validity to the battery, hence a more efficient way of predicting talent through physical fitness is possible. The investigation can benefit sports science because it combines machine learning with practical performance evaluation to ensure that coaches and sport organisations have access to a valid and data-supported tool which enables them to significantly optimize the talent identification process with minimal resource utilisation.