Introduction

Three Chinese and two German players (International Table Tennis Federation, 2018) are in the top five of the current table tennis world rankings (November 2018). To reach the world’s top athletes’ performance level, future generations must be optimally educated and trained in sports. In this process, the search for table tennis talents can be seen as a basic step. Some long-term talent development programs use motor tests as part of a talent screening campaign to assess the characteristics profiles of young athletes (Hohmann, Siener, & He, 2018; Pion, 2015). In a second step, based on the individual level of skills and abilities, a personal recommendation for a certain sport that best matches the athlete’s performance is given to each participant (Pion, 2015, 2017). Since many sports, including table tennis, are accompanied by a high level of technical, coordination and tactical demands (Faber, Bustin, Oosterveld, Elferink-Gemser, & Nijhuis-Van der Sanden, 2016; Munivrana, Furjan-Mandić, & Kondrič, 2015a) that need to be developed early on, many talent search programs start in primary school (e.g., Stemper, Bachmann, Diehlmann, & Kemper, 2009; Pion, 2015; Fuchslocher, Romann, Rüdisüli, Birrer, & Hollenstein, 2011; Golle, Muehlbauer, Wick, & Granacher, 2015). On the one hand, early recognition of talents goes hand in hand with a possible early entry into a suitable sport and, according to Doherty, Martinent, Martindale, and Faber (2018), there is also a higher chance of competing in professional competitions for a greater number of training years. On the other hand, some sports require a very early start, since the age in the international field is also very young (Papic, Rogulj, & Plestina, 2009). According to Hohmann and Seidel (2003), sport recommendations can lead to a higher overall quality in sport clubs. This can be explained by the fact that talents do not “waste” their potential in a sport that is inappropriate for them, but can develop their qualities early in their most suitable sport (Pion, 2015). These sports recommendations made in the context of talent orientation can focus on groups of sports such as game sports (Pion, 2015) or more precisely on single sports disciplines such as alpine skiing (Mueller, Mueller, Kornexl, & Raschner, 2015).

It is sometimes difficult to make a reliable table tennis recommendation as part of the talent search programs. Table tennis is characterized by outstanding motor control (perceptuo-motor skills and coordination) (Faber, Elferink-Gemser, Oosterveld, Twisk, Nijhuis-Van der Sanden, 2017a; Munivrana, Petrinović, & Kondrič, 2015b), fast footwork (Nikolic, Furjan-Mandić, & Kondrič, 2014), timing, as well as anticipatory (Akpinar, Devrilmes, & Kiraczi, 2012) and cognitive skills (Chu, Chen, Chen, Huang, & Hung, 2012; Lopez & Santelices, 2012). A large number of these factors cannot be adequately assessed by the usual motor tests (Bös & Schlenker, 2016). With regard to these difficulties, some authors try to optimize the test selection (Faber, Pion, Munivrana, Faber, Nijhuis-Van der Sanden, 2017b; Faber, Oosterveld, & Nijhuis-Van der Sanden, 2014). The risk here lies in the generality–specificity continuum. If one wants to reproduce the characteristics of certain sports well, the tests become more and more specific and the test battery larger. Furthermore, there is a danger that more techniques than abilities will be tested.

Additionally, to optimize the overall test result, there is also the possibility of weighting the skills in an analysis to obtain a clearer picture of the characteristic test profile (Hohmann, Fehr, Siener, & Hochstein, 2017a) and thus allow a more valid recommendation. By means of differentiated weighting, individual recommendation scores can be determined for each sport and can then be compared. Finally, the highest recommendation value serves as the basis of a future performance forecast. In addition to the “positive” prerequisites, various skills can also seem unimportant or have a negative influence on later success (Kondrič, Zagatto, & Sekulic, 2013). Perhaps it can be seen from these complex recommendation structures why, with a few exceptions (Faber, Elferink-Gemser, Oosterveld, Twisk, & Nijhuis-Van der Sanden, 2017a; Hohmann et al., 2018), there is a large gap in the long-term studies on the prognostic value of childhood performance prerequisites for performance in adolescence. As many studies only consider short or medium time intervals (Höner & Votteler, 2016; Höner, Leyhr, & Kelava, 2017; Le Gall, Carling, Williams, & Reilly, 2010), the results are sometimes ambivalent. While some of these studies could prove the validity of sport-specific motor tests (Höner & Votteler, 2016; Figueiredo, Concalves, Coelho, Silva, & Malina, 2009; Zuber, Zibung, & Conzelmann, 2016), other studies showed no significant correlation between initial test performance and future sport success (Carling, Le Gall, & Malina, 2012; Le Gall et al., 2010). So far, it remains questionable, not least due to the many short-term investigations, whether there is a direct connection between initial performance and later success in certain sports.

The aim of this study is to narrow the gap and look at the prognostic validity of talent orientation using table tennis as an example. The midterm prognostic validity of a general motor test battery and a calculated table tennis recommendation score will be evaluated. To ensure this, competitive players at the table tennis performance levels of U13–U17 (age range from 11.0 to 17.0 years) in the 2016/2017 season (closing date was September 30, 2017) were compared, with their skill profiles collected in a test battery at the U9 age (age range from about 7.0 to 9.0 years) from the years 2010–2014. It was hypothesized that those athletes that performed better in early testing and received a higher table tennis recommendation score (TTRS) should, 4–9 years later (U13–U17), play on a comparatively higher competition level than athletes with lower test scores. In addition, the situation arises where, in single tests of the initial talent screening campaigns, these later table tennis players performed better than their counterparts performed in other sports.

Methods

General study design

The study is based on the results of the Fulda Movement Check (FMC) campaign between 2010 and 2014. The FMC was first launched as a pilot project in 2010 and is a talent and health campaign of the city and district of Fulda together with the University of Bayreuth (Hohmann, Fehr, & Voigt, 2015; see also Hohmann et al., 2017a; Hohmann, Fehr, Siener, & Hochstein, 2017b). This talent screening campaign tested almost all second classes of the Fulda district and city primary schools (more than 13,500 participants) to date by means of a general motor test battery. In addition to general anthropometric parameters such as height and weight, eight general tests of the German Motor Test (GMT) 6–18 (Bös & Schlenker, 2016) and a ball-throw are carried out. Based on the individual results, a recommendation score (RS) is calculated for each of 25 different sports disciplines. The three highest scores then result in the individual sports recommendation, which consequently consists of three sports suitable for each child. At the end of the tests, each participant is handed a certificate showing the different test results and listing the three individual sports recommendations derived from the individual ability profile. The sport aptitudes are documented using a percentage scale. The recommendations focus on medium-to-high performance sports disciplines available in the Fulda region. In Fulda these include, for example, swimming, track and field, gymnastics, cross-country skiing, soccer, handball, water polo, tennis, table tennis, judo, and canoeing. Other sports that do not provide high performance–oriented training (e.g., basketball, badminton, golf, etc.) or where training only begins at a higher age (e.g., baseball, rugby, boxing, etc.) are not recommended. Those athletes that have arrived at a table tennis club after the testing, regardless of their initial recommendation, were determined by daily print media and table tennis website analysis up to the deadline of this study (September 30, 2017). The initial individual test performances in the FMC at age U9 were then compared with the maximum table tennis competition performance up to ages U13–U17, to provide information about the prognostic validity of the early talent orientation.

Participants

A total of 3190 participants of a test campaign between 2010 and 2014 were found to take part in later sports competitions. Of them, 225 students played table tennis (n♀ = 84 and n♂ = 141) until the cut-off date (September 30, 2017). Their mean age at the time of initial FMC testing was M = 94.1 ± 5.2 months. Fig. 1 shows the exact distribution of the participants tested in the FMC 2010–2014. Soccer was explicitly mentioned here, as this sport is the most widespread in Germany, leading to a comparably good data basis (n = 316). Performance until the cut-off date in the 2016/2017 season was individually evaluated for each player using a four-level scale (see Table 1). Level 4 represents the highest performance group of the determined table tennis players.

Fig. 1
figure 1

Scheme of the different participants’ groups of the Fulda Movement Check (2010–2014) included in this study and their performance levels in adolescence (2016/2017)

Full size image
Table 1 Four-level scale recording the participants’ sport-specific competition results of the FMC participants taking part in official table tennis competitions in the age groups of U13 to U17 (N = 225)
Full size table

To generate a larger sample, the test values for the years 2010, 2011, 2012, 2013, and 2014 were analyzed. However, as a result, the players’ ages at the end of the 2016/17 season varies widely between Min = 132 months (11.0 years) and Max = 200 months (16.7 years). For the lowest age difference at the time when the performance levels were collected, this means that the test person had reached his or her highest performance level more than 3 years (3.6 years) after testing in the FMC. The average age difference between the initial performance shown in U9 and the related performance shown in U13–U17 is 5.1 ± 1.0 years.

Within the scope of the FMC campaign, all parents of participating children were fully informed about the evaluations and objectives of the study and provided written consent. University and school ethical guidelines were strictly adhered to.

Measurements

Anthropometric characteristics and motor abilities

In the FMC campaign, anthropometric data were collected in addition to calendar age (in months) and gender. Body mass index (BMI) and Broca index were later calculated from height and weight. The motor tests were based on the German Motor Test 6–18 (GMT) (Bös & Schlenker, 2016). In the test battery, the abilities sprintFootnote 1, coordinationFootnote 2, balanceFootnote 3, flexibilityFootnote 4, arm and upper body strengthFootnote 5, leg powerFootnote 6, and enduranceFootnote 7 are tested. The test was conducted in accordance with the execution methods of the corresponding test manual. In addition to this general assessment of the GMT 6–18, a ball-throwFootnote 8 was performed to show the importance of throwing for many sports (water polo, handball, track and field) and thus to improve the accuracy of the recommendations for these sports. In sit-ups, push-ups, and the 6‑min run, only one attempt was allowed, whereas two attempts were possible in all other disciplines. With the exception of sideward jumping (the average value was counted here), only the best trial was used for further data analysis.

In Bös and Schlenker’s (2016) reliability investigations, nearly 50,000 children in the 7–11 age groups of the GMT 6–18 showed comparatively good values in the test-retest correlation. The coefficient was rtt = 0.82. Balancing backward achieved the lowest value with rtt = 0.52 and bend forward the highest value with rtt = 0.94. The low reliability of the balance test has many causes. One reason lies in the concentration and motivation of the test persons—this is not always completely given in children with repeated tests (see Bös und Schlenker, 2016). Supplemented ball-throwing showed a test-retest reliability of rtt = 0.82 (p < 0.001; n = 3193) in the authors’ own study.

Since age might have an influence on the evaluation of the predictors’ prognostic validity (Meylan, Cronin, Oliver, & Hughes, 2010; Carling & Collins, 2014; Höner et al., 2017; Höner & Votteler, 2016) for all variables, a univariate ANOVA was used to test the data set for significant differences in age. As age was rising with the birth quartiles (F(3; 3186) = 7.61, p < 0.05; Hohmann et al., 2018), the calendar age (in months) was partialized out of the test results by bivariate regression analysis to avoid confounding effects in the following analysis (Willimczik, 1982; Hohmann et al., 2015, 2017a, b). Subsequently, the residuals of this analysis were z‑standardized.

The TTRS was composed of six weighted test results calculated by means of weight factors (WF) in descending order of importance from the sideward jumping (WF = 2.0), balancing backward (WF = 1.8), push-ups (WF = 1.6), standing long jump (WF = 1.4), ball-throw (WF = 1.1), and bend forward (WF = 1.1) values. The weighting allows a better representation of table tennis talent, which is in line with former investigations of elite sport players (see Hohmann et al., 2015). An RS calculated in this way is determined for each of the sports of interest, and thus the children’s primary talents are identified.

Level of sport-specific competition

To compare the performances of the table tennis players with one another and also with athletes of other sports, different competition levels were defined. According to this principle, all players that participated in an official competition before the deadline of the 2016/17 season received a rating from level 1 to level 4 (Table 1). The lowest class were assigned to level 1. These athletes did not receive any notable awards and tended to practice the sport as a hobby (n = 62). The second level indicates that the athletes have played official matches beyond the lowest local level. This performance range contains n = 96 of the tested students. If an athlete performed at regional competitions or championships, he was assigned to level 3 (n = 44). All higher performances above this level were assigned to level 4 (n = 23). This highest level also comprises a female member of the German youth national team in table tennis, thus underlining the high quality of this subgroup. Nevertheless, it is difficult in table tennis to correctly determine a player’s performance despite the existing ranking systems (Gonaus & Mueller, 2012; Roescher, Elferink-Gemser, Huijgen, & Visscher, 2010). To reduce the general difficulties in level assignment, regional talent scouts and table tennis coaches were asked to confirm the ratings.

To investigate the performance of table tennis players, surveys of coaches and active sports clubs’ regional websites were conducted. A large data set was taken from the archives of the local daily newspaper and current local reporting. In addition, the results of the official cooperation partners of the “German Table Tennis Federation” (DTTB, www.tischtennis.de) were retrieved.

Statistical analysis

All analyses and evaluations were performed using SPSS software (version 25.0; SPSS Inc., Chicago, IL, USA), and statistical significance was set at p < 0.05. The participants’ data collected in the years 2010–2014 were related to the test age in months. For this purpose, the calendar monthly age was determined for each student test value, and the age profiles (mean values, standard deviations) were analyzed using a bivariate linear regression. The z‑standardized residuals of these regression calculations of the individual test values were then used for all subsequent evaluations (Hohmann et al., 2015).

To show significant group differences, a univariate ANOVA was performed for levels 1–4. This indicates the prognostic validity of the nine individual tests and the calculated TTRSs. Group differences were analyzed with a post-hoc test. The Bonferroni test was used for analyses of variance equivalence and the Dunnett T3 test for unequal variances.

On the basis of cross-tabulations, odds ratios (ORs) were determined. The OR informs about the chances of reaching the national performance level (level 4) when the test results exceeded a z-value of 1.0. The z‑value of 1.0 determines the limit of one standard deviation and thus corresponds to approximately the best 16% of the students in the individual test results. For each of the nine individual tests and the TTRS, the probability of reaching level 4 was calculated for the best performers compared to the 84% of the weaker children.

To obtain a midterm talent prognosis by means of both a linear discriminant analysis and a nonlinear neural network (multilayer perceptron, MLP), six of the nine FMC juvenile performance prerequisites (according to the TTRS: sideward jumping, push-ups, balancing backward, ball-throw, standing long jump, and bend forward) were chosen to predict two final groups of table tennis players. For this analysis, only the six TTRS tests were used to compare the three results (TTRS, discriminant analysis, and neural network). The two performance groups were formed according to the performance level achieved at early adolescence in the age groups U13 to U17 (see Table 1). Group 1 contained the n = 23 better players from the national level 4, while group 2 consisted of n = 202 players from the lower levels 1–3. To obtain a “true” prognosis, the talent forecasts on the basis of the stepwise discriminant analysis must follow a cross-validation classification (“leave-one-out”). Thus, the total number of all cases minus one was used to compute the discriminant functions that were then used to determine the later performance group of the remaining single case. Similarly, for the MLP analysis, three subsets were created for: (i) training and (ii) testing the predictive model, as well as determining (iii) the final classification of the left-out cases. The architecture of the feed-forward net consisted of the seven input neurons (the six z‑transformed test values: balancing backward, sideward jumping, bend forward, standing long jump, ball-throw, and push-ups and a distortion neuron), one hidden layer with a maximum of four neurons independently selected by the system, and the two output neurons (high performer and low performer). The number of iterations was automatically selected and the process stopped when no further reduction of the error was detected. Since the sample size of the high performers (level 4) was too small for a general statement, these cases were doubled and, together with a group of low performers (levels 1–3) of the same size, used for the analysis. The sample alignment prevented the otherwise larger group of weaker athletes from distorting the neuronal network due to the greater a‑priori probability. From the resulting total of 80 cases, 80% (approximately 64 cases) were used to train the neural network and another 10% (approximately eight cases) to test the resulting network (Hohmann et al., 2018). Finally, the classification was calculated for the hold-out of the remaining 10% of cases (approximately eight cases) that were not used for training and testing the network. This specific type of hold-out strategy was repeated ten times, so that each of the 80 cases should, at least once, be among the left-out athletes finally classified. For each of these ten trials, a new network structure was identified, and independent classification results were produced. To include the remaining 160 cases of low performers in the study, this procedure was carried out on five different randomly selected groups (80 participants each) so that, in the end, about 50 different network structures and thus also 50 different classification results were available. To quantify the validity of this talent identification strategy, the percentage of correct hits of the neural network classifications was averaged over the 50 trials, and the mean value was then used. The classification qualities of both methods (discriminant analysis and MLP) were expressed by the proportion of correct hits, which refers to the percentage of athletes who were correctly assigned as true positives to the group of high performers. An athlete was defined as a false positive if he was classified as a talent but did not reach the best-performing groups. On the basis of the ratio of correct to erroneous hits, the prognostic validity of the test results was calculated.

Results

Test performance results for table tennis players, soccer players, other athletes, and nonathletes

Comparing the test results of later table tennis players, soccer players, other athletes, and nonathletes, it is noteworthy that the tested table tennis players could not deliver outstanding results in any of the tests. It should be noted, however, that this group was the youngest of all the groups tested. Regarding the anthropometric tests, there are no major differences in the results. The descriptive performance characteristics of the FMC participants are presented in Table 2.

Table 2 Descriptive statistics for the two anthropometric and nine motor diagnostics of second-grade students taking part in the FMC 2010–2014 and playing table tennis later on in age groups U13–U17 until the end of season 2016/17
Full size table

The information from the table is not sufficient to elucidate the individual sports’ different characteristics. Differences in gender and test age (55.1% of the tested students are under the age of 95 months) require consideration of the corresponding z‑values to obtain a clearer picture. Fig. 2 demonstrates the early talent screening results for z‑standardized predictors. Compared to soccer players, other athletes from 30 different sports, and nonathletes, the table tennis players reached the best results in sideward jumping and push-ups. The mean z‑values of these two predictors show that table tennis players belong to the best 15% of all tested athletes in these tests. In the sit-up test, the table tennis players reached the lowest value of all tested predictors and were just as good as the nonathletes. Overall, this shows that the table tennis players are weaker in athletic abilities such as endurance and strength compared to the much better soccer players and even to the average of the other athletes. Only in the coordination skills (which also include push-ups) do they show obviously better test results than the other groups.

Fig. 2
figure 2

Fulda Movement Check test performances (z-values) for later table tennis players, soccer players, other athletes, and nonathletes held in second grade at elementary level

Full size image

Prediction of general motor tests

Comparison of different sport-specific competition levels

The comparison of the four groups of table tennis players (Fig. 3) shows that the highest group representing the national level scores best in five of the nine tests. This performance increase could not be detected only in balancing backward, sprint, standing long jump, and endurance run. Within the groups of table tennis players, the ANOVA demonstrated significant differences (p < 0.05) in sit-upsFootnote 9, standing long jumpFootnote 10, push-upsFootnote 11, and endurance runFootnote 12. In push-ups, the post-hoc test (Bonferroni; p < 0.05) pointed out that there are significant group differences between group four (national level) and all other groups. In the endurance run, the Dunnett T3 test (Levene; p = 0.41) also shows significant differences between the lower and higher groups (levels: 1–3; 1–4; 2–3; 2–4). The results for sideward jumpingFootnote 13 and TTRSFootnote 14 showed a tendency to be significant. As the anthropometric characteristics body height and weight did not relate significantly to the later table tennis performance, those data were not presented in the figure.

Fig. 3
figure 3

Prognostic validity of the test performances of the Fulda Movement Check participants in regard to the later competition performance levels at adolescence (U13–U17). Table Tennis … stands for the table tennis recommendation score. Data on the 20-m sprint were transformed in a positive direction

Full size image

Odds ratios for each test and the table tennis recommendation score

ORs can be an effective way of discovering which tests provide a good prediction of young athletes’ future success. The purpose of this study is to use the ORs to determine whether an athlete with a good test score in one of the nine initial tests has a higher chance of reaching the national level (level 4) in adolescence compared to an athlete with a lower test score. For each test, we looked at the chances of reaching the national level at the threshold of a z-value of over 1.0. For example, considering the ball-throwing test, one out of five students with a z-score greater than 1.0 reached competition level 4, whereas only one of eleven of the students with a z-score less than 1.0 played in this league. The chance for a good ball-throwing student to reach the national level is therefore more than twice as high (OR = 2.4) as the chance of a less talented student. In Fig. 4, the ORs for the nine single tests and the TTRSs for those children belonging to the best 16% in this single predictor, which corresponds to a z-value of more than 1.0, were calculated. Thus, the most meaningful test is the push-up test. It shows the highest score with a value of OR = 3.8, whereas (as also shown in Fig. 3) the two predicators sit-ups and balancing backward show values of less than 1.0, which indicates a nonexistent or even negative effect on future success.

Fig. 4
figure 4

Odds ratios (ORs) of the nine Fulda Movement Check (FMC) general motor tests (at the z ≥ 1.0 and PR ≥ 84 threshold) and the more specific recommendation score predict the relative chances of the U9 FMC participants reaching at least the national level (level 4) of table tennis competition performance by their competition age (U13–U17)

Full size image

Looking at the distribution of students within the national-level 4 group, it can be seen that 82.6% of the athletes have a z-score greater than 1.0 in push-ups (Fig. 5). Furthermore, in the sideward jumping test, the percentage of high performers at level 4 is very high (78.3%). From a statistical point of view, the chance of any player reaching a z-value above 1.0 in a test should be about 16%. In this study, the percentages of the national-level players exceed this number in almost all predictors: push-ups (82.6%), sideward jumping (78.3%), ball-throw (39.1%), standing long jump and endurance run (30.4%), bend forward (22.7%), and balancing (17.4%). An exception is the sprint and sit-ups tests, where the percentage is just 8.7% and well below the statistical share.

Specificity and sensitivity

In the context of talent orientation, it is of great interest to know which threshold in the TTRS most likely separates the talents (e.g., national level 4) from the nontalents (e.g., levels 1–3). Depending on the selected limit, the sensitivity and specificity of the TTRS can be determined. In Fig. 6, the respective sensitivities and specificities (U13–U17) for the TTRS score are documented. If, for example, a threshold value of 80% is selected for the U13–U17 players in the TTRS, 16 of the 23 players at the national level can be correctly identified (sensitivity is 69.7%). With the same limit value, 101 of the 202 later low performers can also be classified correctly (specificity is 50%). Based on these results, the threshold at which the aim of the sports recommendation is done best can be judged. For example, the largest sum of the percentages of both values is 121.4% (78.3% versus 43.1%), which is found at a TTRS threshold of 77.5%. For comparison, if every second child is generally recommended as a talent, a 50%–50% result should be obtained. The threshold value found is therefore 21.4% more accurate than a random result. A balanced value of both percentages (59% versus 59%) can be found at a TTRS limit of 81.6%.

Fig. 5
figure 5

Percentage of high performers (z-value ≥1) and low performers (z-value <1) in the nine single tests within the group (U13–U17) of the national level (level 4) table tennis players (the line indicates the statistical average)

Full size image

A cross-validated discriminant analysis can be used to determine a “true” prognosis of future success. Based on the six test values composing the TTRS (sideward jumping, push-ups, bend forward, standing long jump, ball-throw, and balancing backward), 68.2% of the later top performers (level 4) could be identified at the age of about 8 years (Fig. 7; n = 219Footnote 15). The overall prognosis leads to a percentage of 69.9%. The best test in this regard was the sideward jumping, correctly identifying 69.6% of the future national-level players in advance. However, only 57.4% of the subsequently weaker athletes could be classified correctly by this single test. Fig. 6 shows the respective totals of the cross-validated discriminant analysis. Other test combinations than the six above-mentioned tests led to partly weaker overall forecasts and fluctuated between a total of 74.8%–50.7% correct hits. Thus, the discriminant analysis provided the best overall result with a combination of seven test values: sprint, balancing backward, sideward jumping, bend forward, endurance run, push-ups, and sit-ups with 74.8% (75% specificity and 72.7% sensitivity).

Fig. 6
figure 6

Specificity and sensitivity of Fulda Movement Check participants’ table tennis recommendation scores to predict future table tennis performance at national levels (level 4) in adolescence (U13–U17)

Full size image
Fig. 7
figure 7

Discriminant analysis to predict the later U13–U17 table tennis national performance group based on initial Fulda Movement Check performances (each full symbol represents 10 children)

Full size image

If an artificial neural network (Multilayer Perceptron; SPSS 25.0) was used, the cross-validated procedure led to a much higher prediction of 78.4% correct hits. The better talent forecast of this nonlinear mathematical tool is made clear particularly in the talented players group. Here, the MLP significantly increased the number of correct high-performance predictions (86.2% true positives). The number of correct predictions of the table tennis players (67.9% true negative predictions) is comparable with the result of the linear discriminant analysis.

Discussion

Although table tennis is a very complex and coordinative sport (Robertson et al., 2018) that requires visual perception, control of arm and head movements, as well as ball control and visual perception (Faber et al., 2014), it turns out that certain skills, which can be assessed in a general motor test set-up, accurately represent the talent make-up of a table tennis player. If one compares the initial performance of the later table tennis players with that of football players (see also Hohmann et al., 2018), it is notable that table tennis players perform weaker in the tests with high athletic demands (e.g., endurance and strength). Robertson et al. (2018) showed similar results in studies between tennis, badminton, and table tennis players. With the exception of four test results, the average of all performance prerequisites was always lower than that of the footballers and sometimes even lower than the scores of the other athletes. It is generally known that a high level of athleticism and the associated strength training are related to muscular shortening. So, it can be assumed that the better mobility of table tennis players is due to the low proportion of strength training in this sport. This assumption is reinforced by the low mean value in the sit-up test. In the field of sit-up performance, the table tennis players showed the weakest values of all tested athletes and only equaled the “nonsport” participants. If the highest youth national level is considered, in our study, good core muscle strength was associated with a negative outcome in the later performance development (ORs).

The anthropometric values (height and weight) did not differ significantly from the values of other athletes. These findings partly correspond to those of Robertson et al. (2018). In test tasks primarily related to general and gross-motor body coordination, however, the later table tennis players achieved very good scores. This included sideward jumping (Nikolic et al., 2014) as well as push-ups. The push-up test is not fundamentally dependent on arm and upper body strength due to the still good weight–force ratio of the participating children and the complex upper-body coordination required in this test (see Bös & Schlenker, 2016). Rather, there seems to be an upper limitation in the coordinative speed the children exhibit. Surveys of table tennis players showed that many top athletes find learning coordinative movements easy: “I … just did it and it was correct” (Faber et al., 2015, p. 9). The push-up test could perhaps represent this fast learning of coordinative movements. The test tasks push-ups and sideward jumping led to significant differences between the sports groups on the one hand and the table tennis performance groups on the other hand. Also, it was shown at the national level that mainly athletes (more than 75%) with a high z‑value (z ≥ 1.0) have reached this performance level.

It is not surprising that the push-up test showed the highest OR values overall. The probability that one of the best 16% of the participants in this test will reach the national level is 3.8 times higher than for the test participants with a lower test value (z-value <1.0). Other studies with elite table tennis athletes have shown similar results (Mähner, 2018). It is astonishing, however, that the push-up test in the older subgroup of U15–U17 did not provide comparable results and only came second in the ORs (ORpush-up = 2.18). Nevertheless, this can be explained by the fact that, in all four performance groups, the z‑value is on average higher than z = 1.0. The distinction between a high-performance group (z-value ≥1) and a low-performance group (z-value <1) could not cover these differences, since almost all groups performed above that limit value. The situation would be different with a z-value limit of for example z ≥ 1.3. The differentiation now would be much stronger, but this feature must be analyzed more thoroughly on the basis of a greater sample size in further studies. Overall, however, this study found that a higher performance level also increases the share of the best performers in a particular test. Only four tests were identified where higher athletic performances did not have a statistically significant impact on later table tennis performance (U13–U17). In addition to the previously mentioned sit-up values, the bend forward, sprint, and balancing backward performances also did not influence later table tennis success. As the U13 sample investigated in this study was much younger than the much older adolescent table tennis athletes investigated by Hohmann et al. (2015; see also Pion et al., 2015), the different relevance of the change in performance in these four motor tests should be investigated more deeply in future studies (Mähner, 2018).

A sport-specific RS is a useful way to better reflect table tennis performance. In this test, too, the TTRS at least distinguished between top performers and hobby athletes. The probability of reaching the national level for athletes with a high TTRS, however, was comparatively low (ORTTRS = 1.51). The previous TTRS calculation, consisting of sideward jumping, balancing backward, push-ups, standing long jump, ball throw, and bend forward, could not be confirmed in the discriminant analysis. Here, the best results (74.8%) were calculated by the seven predictors: sprint, balancing backward, sideward jumping, bend forward, endurance run, push-ups, and sit-ups. While in the post-hoc test the three variables sit-ups, push-ups, and endurance run significantly separated the performance groups, the sideward jumping values in the ORs showed better quality (ORsideward_jumping = 3.68). This goes hand in hand with the results of Doherty et al. (2018). In spite of the potentially expandable recommendation calculation, the previous set-up in sensitivity and specificity led to a good overview in a wide range of border areas.

Conclusion

Prognostic validity studies on talent screening and orientation campaigns, such as this one, build on the evolving long-term performance development of the tested participants. In addition to the physiological prerequisites (Kondrič, Furjan-Mandić, Kondrič, & Gabaglio, 2010), many studies (Lopez & Santelices, 2012; Chu, Chen, Chen, Huang, & Hung, 2012; Rodrigues, Vickers, & Williams, 2002) focus on mental, cognitive, and perceptual abilities. These studies indirectly demand that talent orientation campaigns must also look at the psychological features of the table tennis sport. Irrespective of such new developments, the present study has already shown that only with the help of test-based and long-term talent test campaigns, such as the Flemish Sport Compass or the FMC, prognostically valid statements can be made about the future success of sports talents. The results of this study contribute to the table tennis talent search and orientation, as they explicitly show the characteristic abilities of successful youth table tennis players at an early age and thus offer a valid strategy to find and identify later talents. For more precise statements in the recommendations, however, it is advisable to integrate further perceptuo-motor tests into the test battery (Faber et al., 2017a) to fairly depict such a complex sport. From this point of view, it may also be necessary to add advanced statistical methods such as neural networks or other nonlinear tools to current talent identification procedures (Pion et al., 2016).