The dataset was composed of French Track and Field athletes aged 12–22 and classified into one-year age categories, from U13 (Under 13) to U23 (Under 23). It included 189,238 male and 169,372 female performances, from athletes who competed at least once in local, regional, national or international competitions since 2009. All recorded performances conformed to standards of the International Association of Athletics Federations and were collected by the French Track and Field Federation. Date of birth, date of competition and all performances were collected. Also, the level of competitiveness was considered as follows: “All,” “Top50%” which corresponds to the best 50% of the age category in the event concerned and “Top10%” which corresponds to the best 10% in an equivalent way. For both male and female, events included 100, 200, 400, 800, 1,000, 400 m hurdles, shot put, discus throw, javelin throw, hammer throw, high jump, pole vault, long jump, triple jump plus 100 m (for female) or 110 m (male) hurdles. Other events were not considered due to lack of data in youth categories.

All statistical analyses were performed using Python (version 3.8.5; Python Software Foundation, Delaware, United States).

Our sample was obtained from the French Federation of track-and-field. All data were reported anonymously. This study was designed and monitored by the IRMES (Institut de Recherche bio-Médicale et d’Epidémiologie du Sport) scientific committee. Data collection was compliant with the General Data Protection Regulations applied in the European Union. The protocol was approved by the ethics panel of CSMF (Conseil Scientifique, Médical et de Formation) from INSEP (Institute National du Sport, de l’Expertise et de la Performance). The use of retrospective data obtained from public sources was also approved by the Commission Nationale de l’Informatique et des Libertés (CNIL No. 2223498), the governing office supervising data utilization in France.

Birth quarters distribution on the youngest age categories for “All,” “Top50%,” and “Top10%” female and male athletes in 100 m sprint is presented in Figure 1. For any sex or performance level, not only the distribution of athletes’ birth quarters significantly differs from the French distribution of births, but also the proportion born in Q4 is lower than the proportion in any other quarter.

For U13 females, the proportion of 100 m sprinters born in Q4 is 21.4%, 19.3% and 16.2% for “All,” “Top50%,” and “Top10%” respectively. Similarly, among males, the proportion of Q1 increases considerably among U14: 27.6%, 34.4%, and 43.4% for “All,” “Top50%,” and “Top10%” respectively.

For males, the RAE is accentuated by the proportion of Q4 among U13 and U17, which are 23.2% and 19.1% respectively in the “All” category. Among the “Top50%” and “Top10%,” the effect is maximal in the U15 category and decreases thereafter. The distribution of birth quarters by age category for all events and sexes for “All” and “Top10%” is presented in the Supplementary Material, with an under-representation of athletes born in Q4 for the majority of events.

After running linear regressions on performance as a function of the relative age gap, the resulting coefficients are presented for 100 m female and male sprint in Tables 1, 2 respectively. The coefficient is highest among the U13, regardless of sex. For females, within the same age group, there can be 564.1 ms (3.875%), 141.0 ms (1.0%), 47.0 ms (.3%) and 1.5 ms (.01%) of difference on average per year, quarter, month and day respectively. For males, the difference is greater, with a year, quarter, month and day generating an advantage of 931.0 ms (+6.5%), 232.8 ms (+1.6%), 77.6 ms (+.5%) and 2.6 ms (+.02%) respectively.

The pre-rebalancing performances for females 100 m for U13–U17 categories are shown in Figure 2A. On the age axis, the majority of the 10-Best performances are concentrated towards the end of the age category concerned, meaning that the most performant are the older even in a same-age category.

The rebalanced performances of the same population according to age and the initial performance is presented in Figure 2B. First, new potentials emerge through this rebalancing method. Indeed, there are 35, 37, 23, 18, and 22 of them, respectively at U13, U14, U15, U16, and U17 to exceed the pre-rebalancing 10-Best performances. Then, among them, rebalanced performances come to feed a new 10-Bestr (10-Best rebalanced). In the youth categories, this 10-Bestr is renewed by almost half and more stable with age.

The principle is similar for Figures 2C, D, with pre-rebalancing and post-rebalancing performances for males 100 m for U13–U17 categories respectively. Again, the best performances before rebalancing are achieved by relatively older athletes. On the other hand, more rebalanced performances are above the 10-Best threshold. In the U15 category, 103 new athletes appear in the list, with great potential.

In order to validate our individual rebalancing method, we compare the rebalanced performance with the second realized performance. The difference between these two performances is shown in Figure 3 in percentage. On average, by category, the rebalanced performance and the second performance achieved are not significantly different. As can be seen, the difference between the two types of performance is centered around zero, so there is no significant difference between the realized performance and the rebalanced performance.

By applying a rebalancing method, many individual potentials that may be lost for the future if undetected are revealed. A quadratic function was applied on 16 and 17 years old male and female athletes (Brustio and Boccia, 2021). Using 16 and 17 years as the reference age, they corrected the performances of the 15.01–16.99 and 16.01–17.99 years old athletes, respectively. After correction, they found a smaller or no RAE among 16 year-old athletes. However, a stronger RAE was revealed among corrected performances of 17 year-old athletes, who ran the 100 m and 200 m. Athletes older than the reference age had their performance deteriorated, while the younger ones had their performance improved. It is then worth noticing that this study was conducted on world-class athletes, who had achieved at least five international performances in their career: it is therefore important to consider the competition level.

7761 sprinters (60 m) between the U9 and U16 categories were considered at different performance levels (Romann and Cobley, 2015). After observing that the RAE increased with performance levels, the authors corrected the performances by applying a regression for all age-groups. They found that RAE was not present anymore for “Top10%,” “Top25%,” and “Top50%” excepted for U9 and U10. This study does not investigate the same age range. As the practice of middle- and long-distances comes at later ages than sprints, we focused on earlier ages for sprints and later ages for other distances. A trend confirmed in a study (Marc et al., 2018), that spotlighted the age difference between the peaks of performance in the 100 m and 6-day ultra-marathon (Marc et al., 2018). The male performance peak was reached at 24.7 years on the 100 m and at 38.1 years for ultra-marathon races. The same trend was observed for females, with a performance peak age at 25 on the 100 m and at 43 for the 6-day ultra-marathon (Marc et al., 2018). This study focuses on different age groups with respect to the considered event.

The main purpose of this study was to reveal the true potential of athletes, once the relative age effect was taken into account. Regressions were run in each age group. Coefficients of determination are high (between .62 and .94 for female, between .81 and .95 for males), showing the consistency of the results and the importance of better appreciating athlete’s potentials with respect to her/his age group. It is now necessary to think in terms of age groups and not over the whole spectrum of age categories.

Correction methods have been applied in youth track-and-field, such as in male and female long jump. In those studies, RAE was spotlight among “All,” “Top25%,” and “Top10%” male long jumpers with a greater effect when the level of competitiveness increased. The authors (Brustio et al., 2022) used polynomial regressions of degree 2 to obtain the rebalancing coefficient. After correction, RAE was no longer spotted among Top25% and Top10% male jumpers. In this study, we chose to run a linear regression by age category to better capture the importance of the relative age effect within each age category. The intention was also to rebalance performances and to observe individually whether this rebalancing allowed potential to emerge. Indeed, the aim was to see the individual impact on performance and not the collective impact on the RAE, hence the rebalancing of all performances and not just the best ones. In addition, validation methods also differ. On one side is checked how much the RAE is reduced by rebalancing the performance. On the other side, the rebalancing method is applied individually to a performance in order to observe how close the rebalanced performance is to the realized one. That is why this method is more revealing of young potentials.

Peak performance is reached earlier in females than in males in these younger categories studying here. This may be related to peak growth velocity. Indeed, studies (Tanner et al., 1966; Largo et al., 1978) showed that girls reach their peak growth velocity between the ages of 11 and 13 years, whereas for boys it happens later, between 13 and 15 years. This has an impact on the rebalancing coefficients, which converge more quickly to 0 than for male.

Also, the popularity of an event may have a non-negligeable impact on RAE (Cobley et al., 2009), especially in popular sports. A concept confirmed by another study, that stated that competition is a fundamental condition for the rise of RAE (Musch and Grondin, 2001).

We can also wonder if the popularity of an event does, or not, provide the conditions for a recurence loop, when more media coverage brings more practitioners and higher competitiveness, reinforcing the process of selection. This may explain our non-significant result of skewed distributions of birth quarters in some age categories. RAE is larger in events where speed and strength are emphasised (Hollings et al., 2014). In the French Youth Swimming Championships (U11 up to U19), for example, the number of female swimmers participating to the Freestyle 1500 m is 25 times lower than in Freestyle 50 m (Difernand et al.). This may explain the link between RAE, competitiveness and popularity.

For all the reasons mentioned above, corrective adjustments should be considered in talent identification and detection programs. Indeed, federations, coaches and staff members should be aware that these methods may reduce biases and dropout rates and increase participation and interest among young talented athletes (Abbott et al., 2020).