The Dutch Technical-Tactical Tennis Test (D4T) for Talent Identification and Development: Psychometric Characteristics

Thirty-two youth male players (age 13.4 ± 0.5; body height 167.7 ± 10.6 cm; body mass 52.3 ± 10.9 kg) were recruited via tennis organizations, clubs and coaches who all gave permission for this study. Participants were classified as elite (n = 15) or sub-elite (n = 17) according to their position on the Dutch national youth ranking list under 15 years at the time of testing (KNLTB). Elite players were those ranked among the top-50, while sub-elite players were classified between position 51 and 350 on the ranking list. Test-retest reliability was assessed in ten subelite players.

The Dutch Technical-Tactical Tennis Test (D4T) consisted of 72 strokes, grouped in four games of six rallies, in which each rally included three strokes. A ball machine (Pro Match Smartshot, Mubo, Gorinchem, Netherlands) was used to meet the criterion of standardization and to feed balls to the participants. Ball accuracy was measured using target areas to which participants were instructed to direct their strokes. A large target area (3.6 × 2.7 m) and a middle target area (2.4 × 1.8 m) were located at both corners at the intersection of the baseline and the sideline as displayed in Picture 1. A small target area (1.2 × 0.9 m) was located inside the middle target area and was positioned 0.45 m from the sideline and 0.6 m from the baseline. The colored target areas were stitched on a large carpet and a colored cap was placed in the center of the small target area as the main target point.

Each game was divided into two offensive, two neutral and two defensive rallies, representing different tactical situations as displayed in Picture 2. Offensive rallies consisted of three ball projections just beyond the service line. Neutral rallies comprised three ball projections to the area around the middle of the court approximately one meter before the baseline, and defensive rallies included three ball projections to the sideline and beyond the service line. The velocity of the ball projections was approximately 70, 75 and 80 km/h, in the offensive, neutral and defensive rallies, respectively. The time interval between the ball projections was 2.5 s in each tactical situation.

The order of the tactical situations (i.e. offensive, neutral and defensive) in games occurred randomly. However, the order of the games during the test was the same for each participant, since the degree of difficulty was increased during successive games. In the first game, during each rally, participants had to return their strokes to the left target area (deuce side). In the second game, they had to direct the strokes to the right target area (advantage side) and finally, in the third game, participants had to aim their strokes alternately between the two target areas. For example, if a participant directed the strokes in one rally to the left/right/left target area, in the next rally they had to aim their strokes to the right/left/right target area. In the final game, the target area was determined by lights which turned red either on the left or right side of the court. The lights were positioned on tripods and placed in both corners of the singles court just behind the baseline and after the target areas. A light gate was placed before the ball machine. After a ball passed the light gate, a signal was sent by an interface and computer to the lights with an adjustable delay set at 500 ms. Following a prescribed protocol and after the signal was given, one of the two lights turned red. The lights were illustrative of the position of an (artificial) opponent. Hence, participants had to return their strokes to the opposite side of the red light. The complete test design is displayed in Picture 2.

Ball velocity was measured using a radar system (Ball coach pocket radar, PR1000-BC) and was recorded after each stroke. In the current study, a high degree of reliability was found between the velocity of the radar system and the velocity calculated using captured video images. The single measure intra-class correlation coefficient (ICC) was .988 with a 95% confidence interval from .979 to .993, F(1, 49) = 162.303, p < .001. Ball accuracy was determined by recording the landing position of the ball using two HD-cameras (JVC Everio GZ-GX1). A total of nine, six and three points were awarded to balls landing inside the small, middle and large target area, respectively, as displayed in Picture 1. When balls landed outside the target areas, although still in the court on the correct side (determined by a given game situation), one point was awarded. When balls landed in the wrong side of the court, outside the singles lines or in the net, zero points were awarded.

A measure of stroke quality, the velocity-accuracy-index (VA-index) was calculated by the following formula wherein velocity was expressed in km per hour (kph): Velocity−accuracyindex=kph2100×sumofachievedpointsnumberofstrockes×9$$Velocity - accuracy\,index = \frac{{kp{h^2}}}{{100}}\, \times \frac{{\left( {sum\,of\,achieved\,points} \right)}}{{\left( {number\,of\,strockes\, \times \,9} \right)}}$$

Due to the nonlinear transformation of ball velocity, the higher the ball velocity, the more a given increment in ball velocity was rewarded, thus the velocity in the formula was squared (kph²) (Vergauwen et al., 1998). The sum of achieved points was described as the amount of points given to balls landing inside the target areas, the number of strokes was defined as the number of strokes in a particular game, tactical situation or complete test.

The study was approved by the ethical committee of the Medical Faculty of the University of Groningen (Groningen, Netherlands, November 19th, 2015) and was consistent with the ethical requirements for human experimentation in accordance with the Declaration of Helsinki. Experimental procedures and potential risks were explained to parents/legal guardians and participants, who all signed a written informed consent form. Two observers measured the players height and sitting height to the nearest mm, and body mass to the nearest 0.1 kg. On testing days, participants performed a 10 min warm-up, including 5 min of groundstrokes. Afterwards, they were alternately tested on an indoor tennis court. In the meantime, the remaining participants conducted a training session at low intensity. During the test, participants were allowed to rest for 20 s in between the rallies and 90 s after three games, which was similar to match play (ITF Tennis). Participants were not limited to a particular stance or grip, but were requested to return balls at their own style (except the use of slice strokes) and match pace. Participants were allowed to use their own racket during the test. Before testing, they played four rallies, one of each different tactical situation (i.e. offensive, neutral and defensive) and one rally with the lights, to get accustomed to the test situation. Test-retest reliability was assessed by ten participants who performed the protocol twice within 14 days.

IBM SPSS Statistics 23 (IBM Corp. Somers, NY) was used for the statistical analyses. Scores on the VA-index, ball velocity, accuracy and percentage errors were checked for normality by exploring normality plots and z-scores for skewness and kurtosis. If values were missing, maximum likelihood estimation was used as substitution method in the missing value analysis. In total 2.1% of the values of the velocity variable were missing. For the interrater reliability, the Cohen’s kappa was used by evaluating the landing position of balls in one session by two observers. The relative and absolute reliability of the D4T were examined using an ICC and a paired t-test, respectively. ICC values were interpreted as poor if < .40, good if ranging between .40 and .75, and excellent if > .75, in

accordance with the reliability level scale suggested by Fleiss (1986). Measures of agreement were determined by the standard error of measurement (SEM), smallest detectable differences (SDD) and a Bland-Altman plot (De Vet et al., 2006; Weir, 2005). The discriminant validity was evaluated by comparing the scores on the VA-index, ball velocity, accuracy and percentage errors of the elite and sub-elite players using independent t-tests. Moreover, two univariate ANOVAs were executed with the performance level as a between-subject factor and the game number (1, 2, 3, and 4) and tactical situation (offensive, neutral and defensive) as within-subject factors for the dependent variable VA-index. These ANOVAs were performed separately, as the VA-index per game number and tactical situation were distinct outcomes which were not possible to analyze together. The Bonferroni post-hoc test was performed when the main effect of the game number or tactical situation on the performance was significant. The concurrent validity was investigated by correlating the VA-index with the national ranking using the Spearman’s correlation coefficient; p-values lower than 5% were considered statistically significant. Finally, the practical feasibility of the D4T was evaluated, although no fixed requirements for feasibility existed (Bowen et al., 2009). Examples of test criteria for good feasibility are such as that the duration of the test should not take too long, the test should be easy to perform by participants and the number of used materials and test leaders should be considerably small.

For the interrater reliability, the Cohen’s kappa revealed an almost perfect agreement between the two observers for the evaluation of the landing position of the balls, k = .98 (p < .001). The results for the test-retest reliability of ten participants showed good relative as well as absolute reliability as presented in Table 1. Figure 1 presents the Bland-Altman plot of the test-retest reliability of the VA-index. The plot shows the difference in the VA-index between the first and the second test-session, a mean difference in the VA-index closer to 0 represents a more reliable measure.

Table 1

Reliability outcomes of the Dutch Technical- Tactical Tennis Test (D4T) in youth tennis players (n=10

Tables 2 and 3 illustrate the mean scores for the VA-index, velocity, accuracy, and percentage errors for the elite and sub-elite players in the D4T in total, and in the different games and tactical situations, respectively. The elite players scored significantly higher than the sub-elite ones on all variables. No differences between the games for the VA-index were found (F(3) = .374, p > .05), and also the interaction between the level of performance and the game number (Level × Game number) for the VA-index was non-significant (p > .05). In contrast, differences were found between tactical situations on the VA-index (F(2) = 9.293, p < .001), players in the offensive situation scored higher on the VA-index than in the defensive situation (p < .001), but not in the neutral situation (p > .05). No differences were found between the VA-index for players in the neutral and defensive situation and also the interaction between the level of performance and tactical situation (Level × Tactical) was non-significant (p > .05).

Table 2

Descriptive statistics (mean ± SD) and differences between elite and sub-elite youth tennis players in the D4T(n = 32)

Table 3

Descriptive statistics (mean ± SD) and differences between elite and sub-elite youth tennis players in different games and tactical situations (n = 32)

The validity for the relation between the VA-index and individual positions on the youth ranking list showed a highly significant Spearman’s correlation coefficient of -.750 (p < .001) as displayed in Figure 2. This means that the players with a high position on the national ranking list tended to have a higher VA-index. The proportion of explained variance was .64, which means that 64% of the variance in the VA-index was explained by the ranking.

The maximum time required to prepare the D4T was approximately 15 min and the mean time required to give clear instructions about the D4T to participants was three minutes. The instructions about the D4T were easy to comprehend. The duration of the D4T was approximately 12 min. The physical demand of the D4T was not very high. In general, the average time required to prepare and administer the D4T to a group of eight participants was three hours. The used materials in the D4T consisted of a ball machine, a radar and light system, cameras and target areas stitched on a carpet. The minimum number of test leaders required to administer the D4T was two.