Author(s): HAMRI, I., Institution: INSEP, Country: FRANCE, Abstract-ID: 1537

INTRODUCTION:
The Relative Age Effect (RAE) refers to an imbalance in birth-date distribution within youth categories, caused by age differences of up to one year between athletes born early and late in the same selection year. In athletics, this bias can influence performance assessment and talent identification. Existing correction methods mainly rely on additive linear models, assuming that RAE affects all athletes similarly. This study challenges this assumption by proposing a multiplicative recalibration model, based on the hypothesis that the effect of relative age depends on performance level.
METHODS:
All performances recorded by the French Athletics Federation between 2004 and 2025 were analyzed in U13 to U16 categories. For each athlete, the best performance per season was retained and linked to date of birth. Only disciplines, seasons and sexes with at least 100 athletes were included.
For each sex, age group and discipline, models were trained on three consecutive seasons and applied to an independent fourth season. Two approaches were compared: (i) a traditional additive correction and (ii) a multiplicative correction in which adjustment depends on the athlete’s performance level.
Model effectiveness was evaluated by examining the relationship between recalibrated performance and relative age in the test samples. Regression slopes were computed for the overall population and for higher-performing athletes (90th percentile). Slopes closer to zero indicate a better correction of RAE.
RESULTS:
A total of 387,568 performances were included, producing 1,247 independent test samples. For mean performance, both models provided similar corrections, with near-zero slopes (additive: 0.05, 95% CI −0.02 to 0.12; multiplicative: 0.09, 95% CI −0.01 to 0.19).
For higher-performing athletes, the multiplicative model reduced residual age-related bias. At the 90th percentile, slopes were lower with the multiplicative model (0.11, 95% CI 0.08 to 0.14) than with the additive model (0.19, 95% CI 0.15 to 0.23), with a significant difference between approaches.
CONCLUSION:
While additive and multiplicative models similarly correct the average RAE, the multiplicative approach provides a more accurate adjustment for higher-performing youth athletes. These findings suggest that RAE is performance-dependent and that correction methods should account for this when used in evaluation and talent identification contexts.