Prediction of 1500 m freestyle performance in age-group swimmers based on synthetic data and symbolic regression.

This study aimed to develop interpretable models of 1500 m freestyle performance by implementing symbolic regression using anthropometric, body composition, biomechanical and physiological variables from original data and derived synthetic data. Twenty-four late- to post-pubertal national-level age-group swimmers participated. Mean comparisons and correlations validated synthetic data fidelity. No differences were observed between real and synthetic data in males or females (p > 0.05). Stroke index and swimming speed (real and synthetic) showed moderate to very strong associations with 1500 m time in both male (r = -0.53 to -0.76) and female swimmers (r = -0.54 to -0.97). In the male model, performance is primarily driven by variables related to stroke efficiency, maximal oxygen uptake, and high-intensity physiological strain. The female model places greater emphasis on pacing-related variables, including mean swimming speed during endurance and [La-] tolerance tests, anaerobic threshold velocity, and metabolic stress indicators. Symbolic regression applied to synthetic datasets is suitable to address the small sample size limitations often present in human research, particularly swimming. Coaches and researchers are encouraged to use symbolic regression and synthetic data and to report the resulting equations to support mechanistic interpretation of 1500 m performance prediction equations for male and female swimmers.