The Explanatory Capacity of Talent Identification Tests for Performance in Triathlon Competitions: A Longitudinal Analysis

Triathlon is a sport that combines the disciplines of swimming, cycling and running, performed consecutively in that order and timed continuously from start to finish without stopping the clock. Between disciplines there is a period known as transition, in which triathletes change their clothing and equipment for the next segment. There are various competition formats; those used in the main world circuit (World Triathlon Series) and in the Olympic Games are sprint (750 m of swimming, 20 km of cycling, 5 km of running) and standard (1500 m, 40 km, 10 km).

As with other endurance sports such as long-distance running and cycling, success in triathlon is related to physiological characteristics of athletes, especially considering high levels in maximum values of oxygen uptake (VO2max) and peak power output in cycling (Wpeak). Similar values to other endurance athletes are also typically observed in submaximal physiological variables, in measurements of the ventilatory threshold (VT) and lactate threshold (LT), although it is difficult to make comparisons due to the multiple definitions that exist (Suriano and Bishop, 2010). Other physiological factors, less studied yet seemingly related to performance, are lactic acid buffering capacity (Baldari et al., 2007) and economy of movement (Millet et al., 2011).

Moreover, some anthropometric variables have been shown to be related to success in competition, such as a low body fat percentage (related to overall performance in competition) and long limbs (arms are related to performance in swimming and legs to performance in running and cycling) (Ackland et al., 1998; Landers et al., 2013). Mechanical factors, such as technical capacity in each discipline (Bentley et al., 2002), and tactical factors, such as competition strategy (Bentley et al., 2002; Hausswirth et al., 2010; Vleck et al., 2008), will also have influence on triathlon performance. In order to analyse these factors, it is important to consider fatigue from the previous disciplines, since behaviour will be different if analysed in isolation (Ofoghi et al., 2016).

The popularisation of triathlon in recent years has led several sports institutions to start investing significant resources in the talent identification process, in order to find young males and females with the potential to become elite athletes in adulthood (Swann et al., 2015). Some institutions, such as the Spanish Triathlon Federation (FETRI), use these programmes to provide the most talented athletes with appropriate training (including greater monitoring through national team training camps and competitions and even scholarships at high performance centres) to enable them to join the Spanish elite team, ensuring high competitiveness in international events (Federación Española de Triatlón, 2008). To achieve this goal it is critically important to evaluate the validity of the models applied and the success rates of the programmes carried out (Vaeyens et al., 2008).

However, there is a lack of scientific research on talent identification processes in sports in general (Johnston et al., 2018), and especially in triathlon. Bottoni et al. (2011) tried to identify factors to be considered in talent identification programmes, suggesting that they should be based not only on measuring physical parameters, but should also take account of other variables. Cuba-Dorado et al. (2015) analysed the talent identification tests developed by the Spanish Triathlon Federation (FETRI), which has one of the highest ITU country rankings worldwide. They found that the tests had very limited explanatory capacity for overall performance in competition in the year the tests were conducted.

This suggests that the tests carried out to choose the most promising Spanish triathletes do not predict their performance. Nevertheless, it seems necessary to check these findings in a longitudinal study, which can help avoid biases in talent identification, reducing the risk of prematurely attributing causality to the results (Johnston et al., 2018). The objective of this study, therefore, was to determine the explanatory capacity of talent identification tests carried out by the FETRI in relation to performance in competition in subsequent years in the discipline corresponding to each test.

In all generalisability analysis models the residual error coincided in the least-squares (VARCOMP Type I) and maximum likelihood (GLM General Linear Model) estimation procedures, as it was confirmed that the sample distribution was normal, linear and homoscedastic (Searle et al., 1992). The precision models of the S100, S1000 and R1000 tests showed high levels of Gr (between 0.99 and 1) and Ga (between 0.93 and 0.99), S1000 being the highest value that the two coefficients could take. However, R400 delivered no result (no comma) with any of the estimation procedures used. The degree of independence was calculated using the Durbin-Watson (D-W) test, in which it could be assumed that a value located between 1.5 and 2.5 showed independence. The values of all the models lay within this range, as can be observed in Tables 1, 2 and 3.

The results of the significant model with the independent variables that showed the greatest explanatory capacity in each of the years analysed are presented below. Those for the R1000 test are shown in Table 1, those for the S1000 test in Table 2 and those for the S100 test in Table 3. No significant model was found for the R400 test.

In Table 1 we can see that the greatest explanatory capacity of the 1000-m running test for split places in running was obtained in the year of the tests. The value of the adjusted coefficient of determination (R²a) (p ≤ 0.001) allowed us to state that 26.2% of the total variance of the run split place in the Spanish Championship was explained by the variables All and Points in the linear regression model, obtaining a moderate model adjustment (Figure 1). It can be expressed as follows:

Figure 1

SCYear1 = 184.060 – 4.207·Points – 1.347·All

In subsequent years this explanatory capacity decreased significantly.

In Table 2 we can see that the greatest explanatory capacity of the 1000-m swimming test for split places in swimming was obtained one year after the tests. The value of R²a (p ≤ 0.001) allowed us to state that 34.0% of the total variance of the swim split place in the Spanish Championship was explained by the variables 4R and Points in the linear regression model, obtaining a moderate model adjustment (Figure 2). It can be expressed as follows:

Figure 2

SCYear2 = 135.779 – 0.804·Points – 1.226·4R

In subsequent years this explanatory capacity decreased significantly.

In Table 3 we can see that the greatest explanatory capacity of the 100-m swimming test for split places in swimming was obtained four years after the tests. The value of R²a (p = 0.007) allowed us to state that 19.5% of the total variance of the swim split place in the Spanish Championship was explained by the variables 4R and Time in the linear regression model, obtaining a moderate model adjustment (Figure 3). It can be expressed as follows:

Figure 3

The main finding of this study is that the explanatory capacity of the tests used by the FETRI in their talent identification process for split performance in triathlon competitions is low (1000 m running test: 26.2%, 1000 m swimming test: 34%, 100 m swimming test: 19.5%).

The results also indicate that the explanatory capacity is related to the length of the test (the most explanatory test was S1000, followed by R1000 and S100, whereas the explanatory capacity of R400 was zero). These results are in line with those obtained by Cuba-Dorado et al. (2015), who analysed the explanatory capacity of the same tests for the overall results in triathlon competitions in the year of the tests.

Explanatory capacity seems to be related to energy demands: the 1000-m swimming test, which has the greatest explanatory capacity, is considered a long-term effort, in which the aerobic pathway is the most used. The 1000-m running test, which has the second highest explanatory capacity, constitutes a medium-term effort, with a balanced use of the aerobic and anaerobic pathways. The other tests, i.e., 100 m swimming and 400 m running, are regarded as short-term efforts with predominance of the anaerobic over the aerobic pathway (Capelli et al., 1998; Gastin, 2001; Spencer and Gastin, 2001).

A reasonable explanation of these findings could be a relationship between predictive capacity and the predominant pathway in the test, which is related, in turn, to physiological performance factors associated with endurance sports. Thus, the S1000 test would be associated with VO2max, VT2, lactic acid buffering capacity and economy of movement, and the R1000 test would be associated with VO2max and also with the capacity to withstand concentrations of hydrogen ions in muscle and blood. On the other hand, to our knowledge the test with greater predominance of the anaerobic pathway does not seem to be related to performance factors in triathlon that have been studied (Baldari et al., 2007; Millet et al., 2011; Suriano and Bishop, 2010). However, these hypotheses still need to be tested, and this would allow us to suggest that a closer relationship with the performance factors of triathlon will help increase the explanatory capacity of talent identification tests.

In relation to the tests with predominance of the anaerobic pathway, greater explanatory capacity was observed in the S100 test compared to R400. Despite the fact that these tests are of very similar duration (62.0 ± 4.1 s for S100 and 62.5 ± 5.7 s for R400), the greater explanatory capacity of the swimming test may be related to the influence of speed over the first metres on triathlon performance (Vleck et al., 2008). Another possible explanation for the low predictive capacity of the talent identification tests in relation to performance factors in triathlon is that these tests are exclusively focused on the physiological component, contrary to the proposals of Abbot and Collins (2004), and do not take account of other factors with a higher influence on triathlon performance, such as anthropometric (Ackland et al., 1998; Landers et al., 2013), psycho-social (Ruiz-Tendero and Martin, 2012), mechanical (Bentley et al., 2002; Ofoghi et al., 2016) and tactical factors (Hausswirth et al., 2010; Ofoghi et al., 2016). Nor do they consider one of the distinctive features of triathlon compared to other long-distance cyclical sports: the influence of the preceding discipline on performance in the following segment (Bentley et al., 2007; Delextrat, 2005; Millet and Vleck, 2000).

The explanatory capacity of FETRI’s tests for predicting performance in the discipline in competition is low. It varies substantially according to the type of the test and the number of years between the test and the competition. However, it seems to be related to the performance factors involved in triathlon.

Coaches and sports managers should reconsider the suitability of the 100 m and 1000 m swimming and 400 m and 1000 m running talent identification tests for elite triathlon performance. It seems necessary to investigate a new model for the talent identification programme, with tests more appropriate to the performance factors of the sport, so as to identify and select triathletes in a more precise way. This would contribute to the better use of the resources spent on these programmes. Moreover, such programmes can substantially determine the development of female and male athletes, by providing facilities for the most outstanding ones, and conversely they can discourage those who do not stand out and lead them to abandon the sport, and it is therefore even more necessary to reconsider the relevance of the identification tests used.