**3. Results**

The results regarding reliability analyses are summarized in Table 1. No systematic bias was present for any of the outcome measures. Relative reliability was good to excellent for the peak torque measures (ICC > 0.80) and acceptable for RTD measures (ICC = 0.5–0.8). Typical errors, expressed as coefficient of variation were low for peak torque measurements for adduction, abduction, and internal rotation, but higher (>10%) for external rotation, flexion, and extension, and for all RTD outcomes. This suggests that the predictive strength of the parameters related to hip strength in this study can be used on the level of a sample with high confidence, while the generalizability to an individual must be done with high caution.

Inter-limb asymmetry values ranged from 0.76% for functional leg length up to 40.35% for RTD during hip flexion (Table 2). Inter-limb asymmetries of hip peak torque were lower (except for internal rotation) than the rate of torque development (3.68–11.52% vs. 10.72–40.35%, *p* < 0.05) (Table 3). Hip extension/flexion (1.32 ± 0.33; 1.30 ± 0.25), abduction/adduction (1.08 ± 0.19; 1.07 ± 0.20) and internal/external rotation (1.13 ± 0.27; 1.18 ± 0.26) ratios were similar (*p* > 0.05) for the left and right side (Table 4). Mean inter-limb asymmetry in peak torque during trunk lateral flexion exceeded the 10% threshold (12.48 ± 9.61 %). Mean inter-limb asymmetry in hip ROM (5.91–13.54%), was above the 10% threshold for extension and internal/external rotations, while the trunk lateral flexion showed good symmetry (2.01 ± 1.27%). Asymmetries in horizontal jumps showed lower results (3.48–4.60%) than various parameters of SLCMJ (4.77–11.12%).


**Table 1.** Reliability results for outcome measures.


**Table 2.** Descriptive and normality analysis of asymmetry variables.

SD = standard deviation, S-W = Shapiro–Wilks tests, PF = peak force, PP = peak power, RTD = rate of torque development, ROM = range of motion.


PT = peak torque, RTD = rate of torque development; SD = standard deviation.


**Table 4.** Hip strength ratio differences between left and right leg (Mann–Whitney U test).

A six-variable regression model explained 48% (R<sup>2</sup> = 0.48; *p* < 0.01) of the variation in the T-test performance (Table 5). T-test time was predicted by left hip internal/external rotation strength ratio (β = −0.58; *p* < 0.01) and inter-limb asymmetries in hip abduction RTD (β = −0.38; *p* = 0.01), hip flexion ROM (β = 0.32; *p* = 0.03), functional leg length (β = 0.31; *p* = 0.02), SLTJ distance (β = 0.29; *p* = 0.04), and peak torque during trunk lateral flexion (β = 0.27; *p* = 0.05).

**Table 5.** Final regression model with six independent variables (dependent T-test).


RTD—rate of torque development; ROM—range of motion; SLTJ—single-leg triple jump.

## **4. Discussion**

The aims of the present study were twofold: (a) to profile elite basketball players in different local strength and range of motion asymmetries of hip and trunk region, global power asymmetries in horizontal and vertical jumping and (b) to quantify the relationship of these asymmetries with COD performance. Results showed different magnitudes of asymmetries among tests, body regions, and parameters.

Regarding the magnitude of asymmetry scores reported in this study, largest asymmetries were found in hip RTD (10.72–40.35%), which were significantly larger (except internal rotation) than peak force asymmetries of different hip action (3.68–11.52%) (as shown in Table 2). Compared to local peak torque asymmetries, rate of torque development showed to be a more sensitive parameter for assessment of asymmetries. That is in accordance with findings of Sarabon et al. [23], who reported that the RTD showed larger magnitudes of asymmetries than peak torque during unilateral isometric knee flexion and extension.

To our knowledge, only one study profiled hip strength ratios in professional athletes using fixed point dynamometer [31]. Although study was done on Australian football players, it detected similar results in flexion/extension mean ratio (0.8) as our study (1.3). Moreover, their mean internal/external ratio was 1.15 which is in accordance to our values (1.13 and 1.18 for left and right leg).

They observed a hip adductor/abductor ratio of 1.05 which is much different from our abduction/adduction ratios (1.08 and 1.07), such differences can be attributed to sport specificity of ball kicking in Australian football.

Our results are also showing various magnitudes in hip range of motion asymmetries (5.91–13.54%), with largest being found for extension (12.80 ± 10.95%), internal rotation (13.54 ± 10.57%), and external rotation (13.00 ± 14.10%). Although some research indicates that there are significant differences in hip ROM between the dominant and non-dominant leg in football players [32], a direct comparison of results is limited because authors have not reported asymmetry indexes.

Mean asymmetry in functional leg length was 0.76 ± 0.62%. To our knowledge, studies that assessed asymmetries in anthropometry had not used the functional leg length to investigate the

relationship with performance instead, they had utilized other anthropometric measurements, such as knee and ankle joint width [33] or lean mass asymmetry [34]. Although, there is no past research to compare our results with, a review of Knutson et al. [35] set a threshold of normal functional leg length discrepancy at 2 cm. The mean absolute asymmetry in our study was 0.9 cm, which is thus considered as normal leg length variation.

Inter-limb asymmetries in vertical jumping parameters (4.77–11.12%) showed larger values compared to horizontal jumps (3.48–4.60%), which is in accordance with research conducted by Lockie et al. [15], who reported SLCMJ mean inter-limb asymmetries at 10.4 % and only 5.4% and 3.3% for horizontal jumps (SLHJ and SLLJ). Similar results were reported by Bishop et al. [36] who found larger inter-limb asymmetries in SLCMJ (12.5%) compared to SLHJ (6.8%). Both of these studies were conducted on football players, which indicates that the variability between the testing methods might be substantially higher than the variability between the athletic populations. With that in mind, it can be suggested that vertical jumping is more sensitive for detecting asymmetries than horizontal jumping.

While all of our maximal strength measures showed an excellent reliability (displayed in Table 1), the explosive strength (rate of torque development) were only moderately reliable. As the structured strength and conditioning training as well as the experience level contribute to reliability of data [37], our data can be interpreted with confidence. In the past years, there has been a lot of debate about defining normal asymmetry threshold, most common one being 10%, but di fferent authors suggested values from 5% to 20% [6]. Taking all that into consideration, variability of asymmetry results in our data shows that asymmetry magnitude is dependent on the specific movement, test and parameter which indicates that a unifying asymmetry threshold cannot be established.

Regression analysis revealed a relationship between asymmetries and performance: six variable model explained 49% of T-test performance variance. Independent variable Beta scores (0.27–0.58) show a small to medium individual relationship between di fferent type of asymmetry and COD performance. Although several studies (including the present study) observed negative influence of asymmetries on COD performance, a number of studies identified contradicting evidence. Lockie et al. [10] showed negative influence (*r* = 0.638, 0.669, *p* < 0.01) of isokinetic concentric (60◦/s, 180◦/s, 240◦/s) and eccentric (30◦/s) knee strength asymmetries and COD performance (assessed with T-test). The study of Coratella et al. [11] reported similar results of association between knee strength asymmetries in slow (30◦/s) and fast (300◦/s) contractions and COD performance (T-test and 180◦ turn test). On the other hand, the relationship between COD performance (180◦ turn test) and strength asymmetries tested with the whole kinetic chain movements (e.g., IMTP) has not been detected [12,13]. This observation could be explained by local strength asymmetries show higher magnitudes. Negative influence of asymmetry in vertical drop jump height and COD performance (180◦ turn test) (*r* = 0.66, *p* < 0.01 and 0.52, *p* < 0.05; depending on the side of the turn) was found by Bishop et al. [21]. Also, using horizontal jumps to assess asymmetries, Madruga-Parera et al. [38] found much lower correlations (*r* = 0.32 and 0.31, *p* <0.05) between asymmetry in horizontal jumping length (SLLJ) and COD performance (V-cut and 180◦ turn test). Such a relationship was not found by Lockie et al. [15], who did not observe significant correlations between asymmetry in vertical (SLCMJ height) and horizontal (SLHJ and SLLJ distance) jumping and COD performance (T-test and 180◦ turn test). Similarly, Loturco et al. [18] suggested that asymmetry in various parameters during single-leg squat jump and CMJ do not influence COD performance (zig-zag test). The study conducted by Maloney et al. [22] is probably the most comparable to ours, it showed that sti ffness and asymmetry in single-leg drop jump explained 63% variance of COD performance (2 × 90◦cut test). However, they used just one type of asymmetry as the secondary predictor, while the sti ffness during drop jump was main predictor in the model.

Our model consists of several independent variables, each representing a di fferent type of asymmetry and together showing a significant relationship with COD performance. This is important because findings indicated an independent nature of asymmetry [39]. The most important finding of this study is the connection between asymmetry and COD performance, but also that testing large

variety of asymmetry types is needed to gain a more complete understanding of athlete asymmetry and its relationship to performance.

Certain limitations of this study should be noted. The main limitation is the modest reliability of hip explosive strength measure, however, we find this acceptable as RTD is a highly variable parameter. Moreover, a slightly larger sample size would have been useful in the linear regression analyses, as the best model included a relatively large number of predictor variables.
