TITLE:
Talent Identification in Youth Sport: A Systematic Review of Multidimensional Predictors
AUTHORS:
Islam Mahmoud
KEYWORDS:
Talent Identification, Youth Sport, Athlete Development, Relative Age Effect, Biological Maturation, Machine Learning
JOURNAL NAME:
Advances in Physical Education,
Vol.16 No.4,
November
13,
2026
ABSTRACT: Discovering athletic talent is one of the most complex predictive challenges in sports science. Unlike assessing adult performance, identifying young talent requires predicting their long-term athletic potential under conditions of biological variability, developmental instability, and environmental inconsistency. Consequently, early detection processes often rely on incomplete performance indicators and are subject to structural biases such as the effects of biological maturation and relative age. This systematic review with an umbrella component aimed to synthesize contemporary primary and secondary evidence (2020-2026) on biological, psychological, and computational indicators used in athletic talent identification systems. In accordance with the PRISMA 2020 guidelines, database searches across major publishing platforms yielded 703 records, of which 32 records met the final inclusion criteria, comprising primary empirical studies, complementary conceptual evidence, and two systematic reviews forming the umbrella component. The results indicate that the selection of athletes in the early stages is strongly influenced by biological maturation and relative age effects, which can distort developmental trajectories and affect the accuracy of talent identification. Recent research increasingly supports multidimensional assessment models that integrate physical, technical, tactical, and cognitive variables, improving predictive accuracy compared to single-domain approaches. In addition, machine learning approaches are emerging as promising analytical tools capable of modeling the complex interactions between talent indicators. However, limitations remain regarding external validation, interpretability, and potential bias in datasets. Overall, the available evidence supports further investigation of hybrid frameworks combining objective assessment, structured expert judgment, and computational modelling. Future research should prioritize longitudinal validation across developmental stages, assessment strategies adapted to maturity, and transparent governance of algorithmic decision-making processes.