Effects of Early Talent Promotion on Junior and Senior Performance: A Systematic Review and Meta-Analysis

The search yielded a total of 51 samples included in 29 study reports from 2009 to 2022. Each study was coded for (1) descriptive data, (2) publication status (published vs unpublished studies, e.g., unpublished theses); (3) sample characteristics (country, sport, sex, age category, performance levels compared); (4) methods of data collection (document analysis or athlete questionnaire or interview); (5) type of TPP (the federation’s squads or youth sport academies); and (6) the effect of the age at commencement of TPP involvement on later performance (see Table S1 in the Electronic Supplementary Material [ESM]).

Table 1 shows characteristics of the total sample. Across studies, all sports of the Olympic Games, as well as cricket, were represented; 68% of the athletes were from team sports (e.g., basketball, cricket, field hockey, handball, ice hockey, soccer, volleyball) and 32% from individual sports (e.g., artistic gymnastics, badminton, fencing, figure skating, judo, race cycling, swimming, tennis, track and field). The participants were from nine countries: Canada, Denmark, Germany, Ireland, Malaysia, the Netherlands, Portugal, Spain, and the UK; 82% were male and 18% female (Table 1).

All analyses were performed using the publicly available R environment, version 4.3.0. Effect sizes are reported as the standardized mean difference (meta-analytic Cohen’s \(\overline{d }\)) of the age at commencement of TPP involvement between higher-performing and lower-performing athletes within a type of sport, age category, sex, country, and type of TPP (federation’s squad or youth sport academy). Effects reported as odds ratios in primary studies were converted to Cohen’s d. Effect sizes were weighted by the inverse within-study error variance of d of each study [64, 65]. Effect sizes (\(\overline{d }\) values) of ~ 0.20, ~ 0.50, and ~ 0.80 were considered as small, medium, and large effects, respectively [66]. We searched for outliers, defined as a Cohen’s d whose residual had a z-score > 3. One outlier was found among the senior samples and was excluded from subsequent analyses ([67], d = 3.56). Partly dependent samples were adjusted using Cheung and Chan’s method [68]. All hypothesis testing was two-tailed; a value of p < 0.05 was considered statistically significant.

The overall effect of the age at commencement of TPP involvement on later performance was estimated by conducting random effect meta-analyses, separately for junior and senior athletes. Then, mixed-effect models with Wald’s F [69] were employed to analyze whether defined subsample characteristics moderated effects. For all moderator analyses, we used the rule of thumb that k ≥ 5 is required for each subgroup [70]. We tested for four moderators: (1) age category: junior and senior athletes; (2) type of TPP: federation’s squad and youth sport academy; (3) performance levels: world class versus lower and national or regional class versus lower; and (4) types of sports: individual versus team sports. There were not enough data (k < 5) for further moderator analyses across more differentiated categories of sports, single sports, or across sexes.

We assessed the quality of the primary studies using the mixed-methods appraisal tool (version 3 [71]). The mixed-methods appraisal tool is a reliable internationally established tool for quality assessment of primary studies in systematic reviews and meta-analyses that provides a dedicated section for quantitative non-randomized studies, as included in this meta-analysis, and has recently been used for similar meta-analyses [63, 72]. We evaluated the primary studies by the two screening questions and the four relevant items for quantitative non-randomized studies within the mixed-methods appraisal tool: representativeness, appropriateness of measurements, completeness of outcome data, and consideration of potential confounders (see Table S2 of the ESM, for further explanation). All studies were assessed by the first author and a random sample of 26 studies (51%) was independently assessed by the second author; inter-rater reliability was excellent (Cohen’s κ = 1.00).

In addition, we examined whether studies conducting data collection by document analysis (e.g., official public roster or squad records) versus athlete questionnaires or interviews differed in effect size. Data were collected by document analysis for 54% and by athlete questionnaires or interviews for 46% of the participants. To investigate potential publication bias, we tested whether publication status (published vs unpublished) was a significant moderator and then inspected the funnel plots and computed Egger’s regression analysis.

Effect sizes did not differ significantly between federations’ squads versus youth sport academies (junior sample: F = 0.615, p = 0.446; senior sample: F = 0.024, p = 0.877), performance levels compared (junior sample: F = 0.285, p = 0.602; senior sample: F = 1.958, p = 0.171), or individual versus team sports (junior sample: F = 0.162, p = 0.694; senior sample: F = 0.002, p = 0.966).

There were not enough data for moderator analyses across single sports. Yet, descriptive data from the analytical categories of sports defined in Güllich et al. [42] suggest consistent results across different types of sports: cgs sports (performance is measured in centimeters, grams, or seconds, e.g., athletics, swimming, race cycling, rowing) juniors \(\overline{d }\) = –0.56, seniors \(\overline{d }\) = 0.36; game sports (e.g., basketball, soccer, rugby, tennis, cricket) juniors \(\overline{d }\) = –0.49, seniors \(\overline{d }\) = 0.55; combat sports (e.g., judo, wrestling, fencing, taekwondo) juniors \(\overline{d }\) = –0.58, seniors \(\overline{d }\) = 0.51; artistic composition sports (e.g., artistic gymnastics, rhythmic gymnastics, figure skating, platform diving) juniors \(\overline{d }\) = –0.25, seniors \(\overline{d }\) = 1.08; and within the largest sport-specific subsample, soccer, juniors \(\overline{d }\) = − 0.53, and seniors \(\overline{d }\) = 0.57 (see Table S1 of the ESM for k and N values).

All primary studies had a high methodological quality and risk of bias was generally low (see Table S2 in the ESM). Effects did not significantly differ between studies using document analyses versus athlete questionnaires or interviews for data collection (F = 2.584, p = 0.118).

Effect sizes also did not significantly differ between published and unpublished studies (junior sample: F = 3.556, p = 0.080; senior sample: F = 0.758, p = 0.390). The funnel plots were widely symmetrical (see Figs. S3 and S4 in the ESM), and Egger’s regression was non-significant (b = 0.88, 95% confidence interval 0.52–1.25, p = 0.149).

The present meta-analysis complements a recent series of meta-analyses [42, 43, 63, 72, 73] that empirically tested the validity of the ‘hard core of assumptions’ [74] of traditional theories of giftedness and expertise [26, 75‐80]. As mentioned above, the assumed premises (1) that achieving a high performance level in childhood/adolescence is a prerequisite for the long-term attainment of a high level of eventual senior performance, (2) that starting sport-specific practice at a younger age leads to a higher level of eventual senior performance, (3) that accumulating a larger amount of organized coach-led main-sport practice leads to a higher level of eventual senior performance, and (4) that accelerated childhood/adolescent performance progress leads to a higher level of eventual senior performance, have all been revealed to be at odds with the empirical evidence [42, 43, 63, 72, 73]. Likewise, the present meta-analysis empirically counters the assumption (5) that younger TPP involvement leads to a higher level of eventual senior performance. Taking the present study and recent evidence together [42, 43], predictors of rapid junior performance and of long-term senior performance are opposite in five aspects: starting age; amount of coach-led main-sport practice; amount of coach-led other-sports practice; age at commencement of TPP involvement; and age of ‘milestone’ achievement.

To illustrate for this discussion, typical performance trajectories of athletes who commence TPP involvement at younger versus older ages are schematically depicted in Fig. 2. Generally, TPPs include the selection of youth athletes and their ‘treatment’ by applying TPP measures to them. The present findings may thus be attributable to specific selection or intervention effects of TPPs, or an interplay of both. Many of the youth athletes selected at a young age have an early biological maturation (onset of puberty and growth spurt), are relatively old within their birth year, and have previously accumulated large amounts of specialized main-sport practice, with little or no other-sports practice [34‐43]. When accelerated early performance progress rests on these factors, this early progress is often associated with reduced long-term sustainability, in that the performance trajectory subsequently flattens (Fig. 2) [34, 36, 42‐49].

Once selected, the TPPs seek to further accelerate the youth athlete’s childhood/adolescent performance progress via expanded sport-specific practice and competitions, supported by corresponding environments, resources, and interventions [7, 9, 16, 22‐25]. The strategy may further boost the youth athlete’s current rate of performance progress and lead to increased short-term performance during junior age categories (Fig. 2). However, it may further compromise long-term sustainability because the large amount of childhood/adolescent specialized practice leads to reduced efficiency of practice in subsequent years—the larger the amount of previously accumulated specialized practice the smaller generally the subsequent performance gain per added practice [42, 43]. In economic terms, these athletes’ diminishing return (Gossen’s first law: diminishing performance improvement per added practice amount [81]) is more pronounced than among peers who had only moderate childhood/adolescent main-sport practice combined with other-sports practice and competitions. The data of Barth et al. [43] have shown that senior world-class athletes began to have an advantage in efficiency of practice over senior national-class counterparts in late adolescence and this advantage in efficiency of practice peaked in athletes’ early 20s.

Furthermore, the expanded time demands, and cumulative physical load associated with early TPP involvement, impose additional opportunity costs on the youth athlete and increase their risks of later occurrences that hinder or end the athletic career (e.g., conflicting time demands from sport and school, declining academic performance, injury, burnout, dropout) [7, 9, 50‐61, 82, 83].

In contrast, many eventual senior world-class athletes developed outside the TPPs, just based on the youth sport program of their home/local sport club or school, until older ages, thus remaining unaffected by potential dysfunctional effects of early TPP involvement. Senior world-class athletes’ childhood/adolescent investment pattern was typically characterized by reduced intensity (less main-sport practice), diversification of investments (two to three sports), greater personal, private contribution (later TPP support), longer total period of “product development” and longer deferral of reward (later ‘milestone’ achievement), reduced total childhood/adolescent costs and long-term risks (injury, burnout, dropout), yielding increased long-term benefit in terms of senior performance (see [42, 43, 84], which also include further in-depth discussion of explanatory hypotheses).

Taken together, the recent and present evidence suggests several clear practical implications.

The study had several strengths, such as a large international sample including a broad range of individual and team sports; considering the most common types of TPPs; distinguishing short-term effects on early junior performance and long-term effects on later senior performance; comparing higher-performing and lower-performing athletes across regional, national, and international levels; comparing athletes within the same age category, type of sport, sex, country, and regarding age at entry into the same TPP, respectively; and a high quality of the primary studies. Nevertheless, several limitations should be acknowledged: (1) the correlational design of the primary studies does not allow for causal conclusions; (2) most athletes were male, from Olympic sports, and from Western industrialized countries (but not from the largest Western country, the USA). Effects of early TPP involvement may differ in Paralympic sports or in different sport systems such as the USA, developing countries, Eastern Europe, China, or Russia. (3) All athletes competed at a regional, national, or international level, which may imply restriction of range. Effects of early TPP involvement may differ among lower-level or more heterogeneous populations. (4) There were not enough data for moderator analyses across single sports. Effects may vary between single sports. (5) Successful senior athletes who were not selected for a TPP during junior age categories may not have been considered in primary studies. (6) The synthesized primary studies used linear methods of data analysis and did not consider non-linear analyses. Finally, although we used multiple databases, as in any systematic review, bias of availability, country, and language is possible.

A goal for future research is to investigate the extent to which opposite effects of the age at commencement of TPP involvement on short-term and long-term performance are due to specific selection or intervention effects of TPPs, or an interplay of both. That research should examine the effects of the individual and combined TPP measures applied to participants on their short-term and long-term performance, while also considering effects on other short-term and long-term outcomes such as athletes’ health, psychosocial well-being, academic achievements, and persistent sport engagement. Furthermore, researchers should seek to extend investigations into TPPs to populations that are under-represented in present research, especially female athletes, more sport-specific samples, in particular from Paralympic and non-Olympic sports, and samples from more countries, in particular the USA, developing countries, East European countries, China, and Russia. In this context, the present findings indicate that long-term TPP effects on eventual senior performance (and perhaps other long-term outcomes) cannot be inferred by extrapolating findings from junior samples [7, 19, 61, 87, 88]. Rather, relevant statements require comparison of higher-performing and lower-performing senior athletes regarding the TPP environments, resources, and intervention measures they were provided during childhood and adolescence.

It is also interesting to note that multiple decades of extensive research into talent identification (TID) [5, 15, 27, 29, 89‐99] is contrasted by lacking research into the purpose TID is done for: i.e., the effects of the TPP interventions applied to the talent-identified TPP participants [16, 22, 23]. The practical use of further expansion of applied research into early TID may be questionable: First, the long-term prognostic validity of the talent indicators worked out through six decades of research has remained poor [5, 22, 27, 34, 94, 97]. Second, because of the generally low “base rate” (the proportion of youth athletes within the population of interest who eventually become successful senior athletes) [18, 22, 100, 101], even substantial improvement of the predictive accuracy of TID would only yield minimal improvement of the “hit rate” of talent selection procedures in practice (the proportion of selected youth athletes who later become successful senior athletes) [see [22] for an exemplary calculation based on empirical data]. Third, perhaps most importantly, early TPP involvement, for which early TID is done, is negatively correlated with long-term senior performance.

In a broader context, it is apparent that there is no single factor that “makes a champion,” but that talent development is multi-factorial, calling for more investigations considering multivariable interactive predictor effects (e.g., [85, 102]). Furthermore, many factors are likely non-linearly related with performance. For example, eventual senior performance is supposedly a parabolic function of several childhood/adolescent predictors: too early or too late start; too much or too little main-sport practice; too much or too little other-sports practice; and too early or too late TPP involvement are supposedly associated with reduced later performance, while there is presumably some respective optimum in between that is associated with increased later performance. Exploring those optima calls for more non-linear multivariable analyses of factors such as machine learning approaches, logistic regressions, or cluster analyses (for recent examples see e.g., [103‐107]).

Finally, the economic concepts of efficiency—performance improvement per magnitude of investment, for example, in terms of amounts of practice and TPP measures and resources—and sustainability—considering the benefit/cost × risk ratio of TPP involvement at shorter and longer terms—generally provide a fruitful heuristic for research into talent development [42, 43, 84, 108] because (1) resources are limited (the athlete’s time, body, load tolerance, and health; coaching and facilities; TPP measures); (2) athletes, coaches, and parents seek to expand benefits for the youth athlete while limiting their costs and risks; and (3) effects of several factors vary and may even be opposite regarding short-term versus long-term outcomes. The concepts of efficiency and sustainability apply to research into TPPs as well as other programs such as youth sport programs in general; coaching; athletes’ participation patterns; the ‘microstructure’ of practice; and athlete services such as physiotherapy, nutritional counseling, or psychological support, and lead to three critical research questions [84]: