Regression Models of Sprint, Vertical Jump, and Change of Direction Performance

Abstract

Swinton, PA, Lloyd, R, Keogh, JWL, Agouris, I, and Stewart, AD. Regression models of sprint, vertical jump, and change of direction performance. J Strength Cond Res 28(7): 1839–1848, 2014—It was the aim of the present study to expand on previous correlation analyses that have attempted to identify factors that influence performance of jumping, sprinting, and changing direction. This was achieved by using a regression approach to obtain linear models that combined anthropometric, strength, and other biomechanical variables. Thirty rugby union players participated in the study (age: 24.2 ± 3.9 years; stature: 181.2 ± 6.6 cm; mass: 94.2 ± 11.1 kg). The athletes' ability to sprint, jump, and change direction was assessed using a 30-m sprint, vertical jump, and 505 agility test, respectively. Regression variables were collected during maximum strength tests (1 repetition maximum [1RM] deadlift and squat) and performance of fast velocity resistance exercises (deadlift and jump squat) using submaximum loads (10–70% 1RM). Force, velocity, power, and rate of force development (RFD) values were measured during fast velocity exercises with the greatest values produced across loads selected for further analysis. Anthropometric data, including lengths, widths, and girths were collected using a 3-dimensional body scanner. Potential regression variables were first identified using correlation analyses. Suitable variables were then regressed using a best subsets approach. Three factor models generally provided the most appropriate balance between explained variance and model complexity. Adjusted R² values of 0.86, 0.82, and 0.67 were obtained for sprint, jump, and change of direction performance, respectively. Anthropometric measurements did not feature in any of the top models because of their strong association with body mass. For each performance measure, variance was best explained by relative maximum strength. Improvements in models were then obtained by including velocity and power values for jumping and sprinting performance, and by including RFD values for change of direction performance.

Introduction

The ability to effectively sprint, jump, and change direction is believed to impact substantially on success in a wide variety of sports (^12,16,21). As a result, numerous studies have sought to identify the factors that determine these abilities, with a view to developing more effective training programs (^2,21,30,37). Previous studies have typically used correlation analyses to identify associations between biomechanical variables and athletic tests that provide a measure of performance. Most frequently, popular resistance exercises such as the squat (^2,6,27), jump squat (^2,33), and power clean (^2,4,21) have been used to collect the input data. Of all variables studied, the relationship between maximum strength and athletic performance has been investigated most often using data collected primarily from 1 repetition maximum (1RM) squat tests. Using a sample of elite soccer players, Wisloff et al. (³⁷) reported correlations of 0.78, −0.71, and −0.68 between absolute 1RM values and vertical jump height, 30 m sprint time, and 10 m shuttle time, respectively. Peterson et al. (³⁰) obtained stronger correlations between maximum strength and similar measures of performance (r = 0.78–0.92) when studying a more heterogeneous group of male and female college athletes. Research has also demonstrated that the relationship between maximum strength and measures of athletic performance may be improved if strength values are normalized relative to body mass (BM) (^27,30).

Based on the hypothesis that muscular power is a key factor determining successful performance in many athletic tasks (¹¹), a large number of correlation studies have measured biomechanical variables during explosive resistance exercises performed with submaximum loads (^2,21,23,33). Many of these studies have also attempted to correlate measures of performance with variables such as velocity, impulse, and rate of force development (RFD). A range of moderate to strong correlations have been reported depending on factors such as exercise selection, methods used to calculate the biomechanical variables, and the sample investigated (¹¹). Hori et al. (²¹) measured power production during performance of a jump squat with an absolute load of 40 kg in a group of Australian Rules football players. The power values demonstrated positive relationships with performance in sprinting, jumping, and quick changes of direction (r = 0.49, 0.54, and 0.39, respectively). Baker and Nance (²) also reported similar moderate correlations between power output and sprinting ability with Rugby league players, but only after power values were normalized relative to BM (r = 0.52–0.61). In general, performance correlations with biomechanical variables collected during explosive resistance exercises have not been as large as those established for maximum strength measures.

It has been suggested that a preoccupation with correlation studies has limited understanding of the variables that best explain performance of sporting tasks (¹¹). Importantly, correlation studies only consider the isolated effect of single variables, whereas performance of sporting tasks may be better explained by combining multiple variables. Instead, it has been suggested that future studies should adopt regression approaches to produce models that combine anthropometric measures with biomechanical variables as it is likely that both phenotypic and force related capabilities impact on performance (¹¹). Additionally, a multiple regression approach has the advantage of increasing the explanatory power of a given model while providing an understanding of how the best variables fit together (³⁵). There have been previous investigations that have incorporated strength (³) or anthropometric variables (¹³) in a multiple regression approach to model performance of common sporting tasks; however, to the best of authors' knowledge, there have been no studies that have included both sets of information in combination with variables believed to reflect important features of the force-velocity and force-time curves. Therefore, the purpose of this study was to expand on previous correlation investigations and identify which combination of strength, anthropometric, and biomechanical variables could best explain performance in sprinting, jumping, and change of direction tasks in well-trained rugby union athletes. The regression approach was not implemented in attempts to predict performance, but rather to model the relationships and improve understanding of the main sources of influence and how these progresses with the addition of further variables. It was hypothesized that the best models would generally combine all 3 categories of measurement. The information obtained from this analysis will assist strength and conditioning coaches in their choice of training practices to selectively target the most important physical and phenotypic factors. In addition, the information could also assist in talent identification.

Methods

Experimental Approach to the Problem

A regression-based approach was used to obtain linear models of performance in common sporting tasks. The explanatory variables of the models included maximum strength values, anthropometric data, and biomechanical measures collected during resistance exercises performed with submaximum loads. The performance outcome variables provided a measure of the athletes' ability to sprint, jump, and change direction. The regression-based approach was used not to obtain specific mathematical formulae or predict performance, but rather to model the relationships and increase understanding of how maximum strength, anthropometry, and biomechanical variables such as power output combine to explain performance in common sporting tasks.

Testing was conducted on 3 separate occasions, with a minimum of 48 hours between each testing occasion to minimize the potential effects of fatigue. On day 1, participants performed sprint and change of direction tests in an indoor gymnasium. On day 2, participants performed maximum strength tests and were assessed for anthropometric characteristics using a 3-dimensional body scanner (Hamamatsu Photonics, Hamamatsu City, Japan). On the final day of testing, participants reported to the human performance laboratory to complete vertical jump tests and perform explosive resistance exercises with submaximum loads.

Subjects

Thirty well-trained nonprofessional male rugby union players (age: 24.2 ± 3.9 years; stature: 182.4 ± 6.7 cm; mass: 94.1 ± 12.3 kg; resistance training experience: 7.3 ± 2.1 years) volunteered to participate in this study. Subjects were recruited from a single team competing in the Scottish Rugby Union Premier League. Each of the athletes regularly performed explosive and maximum resistance exercises as part of their strength and conditioning regime (average frequency of 4 d·wk⁻¹). In addition, the athletes regularly performed sprint, vertical jump, and change of direction tests as part of their ongoing physical assessment. The study was conducted 6 weeks into the athletes' preseason training after a deload microcycle. Subjects were notified about the potential risks involved and provided written informed consent to be included in the study. Previous institutional approval was obtained by the ethics review panel at Robert Gordon University, Aberdeen, United Kingdom.

All testing sessions were conducted at the same time of day to correspond with the athletes' regular training schedule. Additionally, this approach enabled the athletes to maintain their habitual nutrition and hydration strategies to limit potential confounding effects.

Day 1: Sprint and Change of Direction Assessment

Timing gates were placed at the start, 5, 10, and 30 m lines to record 3 different sprint times representing distinct qualities (first-step quickness, acceleration, and speed, respectively) (¹²). All athletes performed a thorough warm-up, which included jogging, dynamic stretches, and a series of submaximum sprints. Athletes started each sprint by adopting a 2-point crouched position, 30 cm back from the starting line. Two maximum sprints were performed with the best time for each split selected for further analysis. Intraclass correlation coefficients (ICCs) obtained from 5, 10, and 30 m sprint times were 0.78, 0.89, and 0.95, respectively.

The ability to change direction was assessed using the 505 agility test (³²). A 15-m track was outlined with a start and stop-timing gate placed at the 10-m line. Athletes sprinted from the start to the end of the 15-m track, where they then turned 180° and sprinted back past the timing gate. Two trials were performed with change of direction made with the left foot, and 2 trials with the right foot. The best time from the 4 trials was selected for further analysis. Intraclass correlation coefficient obtained from the 2 fastest times was 0.80.

Day 2: Maximum Strength and Anthropometry Assessment

On day 2 of testing, the athletes first reported to the gymnasium where they performed 1RM tests in the back squat and deadlift in a randomized order. The athletes were accustomed to performing multiple 1RM tests in a single session as part of their regular physical assessments. To minimize the likelihood of fatigue influencing results, a 30-minute rest period was allocated between tests. Based on a predicted 1RM load, subjects performed a series of warm-up sets and up to 5 maximum attempts. A minimum of 2 minutes and a maximum of 4 minutes recovery time was allocated between each attempt. Within this time frame, subjects self-selected when to perform the lift based on their own perception of when they had recovered. For both the squat and deadlift, a lift was deemed successful if the barbell was not lowered at any point during the ascent phase and on completion of the movement, the body posture was held erect with the knee and hip fully extended.

Anthropometric measurements were made using a Hamamatsu Bodyline scanner that used a Class I laser (eye-safe) device. A total of 15 anthropometric measurements were made, including BM, lengths (stature, trunk, floor to hip, thigh, lower-leg), widths (shoulder, chest), and girths (chest, waist, upper arm, forearm, hip, thigh, calf). Two scans of approximately 10 seconds in duration were made for each athlete. The laser rangefinder created a pixel point cloud representation of the body surface. From these data, proprietary software rendered a polygon shell that could then be graphically shaded for viewing and inspection. Using the high-resolution scanner mode, which samples using a vertical pitch of 2.5 mm, the image created (Figure 1) is an accurate model, which can be viewed as a solid object wire-frame mesh. Intraclass correlation coefficients obtained from the anthropometric variables ranged from 0.95 to 0.99.

Figure 1: Rendered polygon shell used to measure linear anthropometric measurements.

Day 3: Vertical Jump and Explosive Resistance Exercise Assessment

On reporting to the human performance laboratory, the athletes performed a thorough warm-up that included jogging on a treadmill, dynamic stretches, and performance of a series of submaximum jumps. Once suitably prepared, the athletes performed 2 maximum vertical jumps with arms held stationary at the side of the body. The jump that resulted in the greatest vertical displacement was selected for further analysis. The athletes then performed maximum velocity deadlifts and jump squats using loads of 10, 20, 30, 40, 50, 60, and 70% of their previously determined 1RM's. Two repetitions were performed for each condition in a single set, with a minimum 2-minute rest period allocated between conditions and a longer rest period (up to 4 minutes) made available if the athlete felt it necessary to produce a maximum performance. All jumps and loaded resistance exercises were performed with a separate piezoelectric force platform (Kistler, Type 9281B Kistler Instruments, Winterthur, Switzerland) under each foot capturing vertical ground reaction force (VGRF) data at 1200 Hz. Force plate data were filtered using a fourth-order, zero-phase lag Butterworth filter with a 50 Hz cutoff. Displacement, velocity, and power data were calculated at the athlete's centre of mass (COM) during unloaded jumps and at the system COM (athlete + external load) during loaded conditions. The kinematic and kinetic variables were calculated using the VGRF-time data and a forward dynamics approach that has been reported previously in the literature (^20,23). Briefly, trials were initiated with subjects standing erect and motionless. Once data acquisition was initiated, subjects performed the eccentric component of the movement to the required depth and then accelerated upwards as fast as possible. Changes in vertical velocity of the COM were calculated by multiplying the net VGRF by the intersample time period divided by the system mass. Instantaneous velocity at the end of each sampling interval was determined by summing the previous changes in vertical velocity to the preinterval absolute velocity, which was equal to zero at the start of the movement. The position change over each interval was calculated by taking the product of absolute velocity and the intersample time period. Vertical position of the COM was then obtained by summing the position changes. Instantaneous power was calculated by taking the product of the VGRF and the concurrent vertical velocity. Rate of force development was calculated from the slope of the VGRF-time curve extending from the transition between eccentric and concentric phases to the maximum value of the first peak. Jump height for unloaded vertical jumps was calculated using constant acceleration equations and the vertical velocity of the system at take-off (²⁵). For the deadlift and jump squat, the largest force, velocity, power and RFD values measured across the submaximum loads were selected for further analysis. Intraclass correlation coefficient values obtained for these variables and the vertical displacement of unloaded jumps ranged from 0.94 to 0.98.

Statistical Analyses

Based on findings from previous research demonstrating that force and power are related to BM (⁸) and that RFD is related to the peak force value obtained (¹), normalized values for these variables were included in addition to absolute values measured. Allometric scaling using an exponent of 0.67 was used to normalize force and power values relative to BM (e.g., average power per BM^0.67) (²²). Simple ratio scaling was used to normalize RFD values relative to the peak force of the slope from which RFD was calculated. Pearson's correlation coefficients were used initially to quantify relationships between anthropometric and biomechanical variables with performance measures. Suitable variables were then regressed using a best subsets approach (MiniTab 15) to create 2 separate models. The first model included scaled maximum strength measures plus suitable anthropometric and biomechanical variables collected during the deadlift. The second model included the same strength and anthropometric variables combined with biomechanical variables collected during the jump squat. Two separate models were used to assess whether the exercise used to collect the biomechanical variables had an influence on the results. The fit of each model was assessed using the adjusted R² value and Mallow's Cp statistic. To assess whether the addition of a variable resulted in a significant improvement on the previous model, the variance ratio of the change in regression sum of squares relative to the mean square residual was assessed under the appropriate F distribution. Sapiro-Wilk tests for normality supported the assumption of a normal distribution. Residual analysis also supported a linear model as a good fit for the data, whereas low Cook statistics demonstrated that models were not unduly influenced by a small number of data points.

Results

Average values for performance measures, maximum strength scores, and basic anthropometry are displayed in Table 1. Relationships between performance measures are presented in Table 2. Significant correlations were obtained between all performance measures illustrating some degree of relatedness between the ability to jump, sprint, and change direction. The largest correlation values were obtained between vertical jump, 10 and 30 m performance, with values remaining significant but decreasing in magnitude when testing associations with 5 m and 505 agility performance. A wide range of correlation values were obtained between anthropometric and performance variables (Table 3). All girth measurements exhibited significant (p ≤ 0.05) negative correlations with performance; that is, with increasing muscular girths, vertical jump, sprinting, and change of direction performance decreased. A similar relationship was obtained for shoulder width, chest width, and vertical jump performance. However, all correlations between performance measures and anthropometric variables became nonsignificant after statistically controlling for the effects of BM. No other significant relationships were obtained between length or width measurements and performance scores.

Table 1: Anthropometric, strength, and performance results (mean ± SD).^*

Table 2: Relationships between performance variables (r values).

Table 3: Relationships between performance and anthropometric variables (r values).

The strongest correlations with performance measures were obtained for maximum strength scores scaled relative to BM (Tables 4 and 5). In general, absolute force, power, and RDF values exhibited small nonsignificant correlations with performance. However, once these variables were normalized relative to BM and peak force, respectively, the strength of the correlations increased with the majority of relationships reaching statistical significance (p ≤ 0.05, Tables 4 and 5).

Table 4: Relationships between performance and biomechanical variables collected during the deadlift (r values).^*

Table 5: Relationships between performance and biomechanical variables collected during the jump squat (r values).^*

Two separate best subsets regression models were developed for each performance measure (Table 6). The first set of models included normalized strength measures, anthropometric variables, and normalized biomechanical variables collected during maximum speed deadlifts. The second set of models included the same strength and anthropometric measurements as used in the first set, but included normalized biomechanical variables from the jump squat instead of the deadlift. In general, similar results were obtained for both sets of models, with those comprising 3 explanatory variables generally providing the most appropriate balance between explained variance and model complexity. In 3 of the 10 models presented, the addition of a third variable failed to improve on the previous 2-factor model, however, in each instance the F value obtained was approaching significance. The addition of a fourth variable in all cases failed to significantly improve the fit of the model. The greatest amount of performance variance could be explained in the 30-m sprint, followed by the vertical jump, 10-m sprint, 505 agility test, and 5-m sprint (Table 6). For both sets of models, performance was best explained by combining normalized maximum strength scores and biomechanical variables rather than anthropometric measurements. Performance in the vertical jump was best explained by an athlete's maximum strength capabilities and their ability to develop high velocities, whereas, performance in the 5-m sprint and 505 agility tests were best explained by maximum strength scores and RFD. Regression models for 10 and 30 m sprints featured primarily maximum strength scores and mechanical power.

Table 6: Best single-, 2-, and 3-predictor regression models for performance measures combining anthropometric, maximum strength, and biomechanical variables collected during the deadlift or jump squat.^*

Discussion

The results of this study demonstrate that a large amount of variance in performance of movement tasks common to most sports can be explained by an athlete's relative maximum strength and their ability to produce high outputs in certain biomechanical variables. Using the adjusted coefficient of determination, between 70 and 80% of the variance in vertical jump, 10-m sprint and 30-m sprint performance could be explained by relatively simple 3 factor models. For the 505 agility and 5-m sprint tests, the explained variance decreased to between 40 and 65%, indicating that factors other than those assessed in the present study are important in determining overall performance. Future models may require inclusion of other fitness measures or technique-related factors to increase understanding of performance, particularly for the acceleration and change of direction tasks. However, the higher within-individual variability measured in the 505 agility and 5-m sprint tests may also have contributed to reduced explanatory power of the models. Correlations between the performance measures were all significant highlighting relatedness in the performance of sprinting, jumping, and change of direction tasks. Again, the magnitudes of the relationships were lowest for the 505 agility and 5-m sprint tests suggesting that performance in these tasks is more distinct and potentially more complex.

The results of the present study support the findings from a number of previous investigations demonstrating strong relationships between maximum strength and measures of athletic performance (^2,30,37). Data collected from the most recent studies have also shown that the strength of the relationship can be enhanced by normalizing maximum strength values relative to BM (²⁷). However, a range of factors including the variation in the sample's athletic capabilities and BM are likely to impact on the strength of relationships obtained. Peterson et al. (³⁰) investigated the relationship between 1RM back squat values and performance in jumping, sprinting, and agility tests. The subjects comprised a heterogeneous group of male and female college athletes from a wide variety of sports. When modeling the entire group (n = 55), the relationships between maximum strength and performance measures were strong and relatively unaffected by normalizing relative to BM. When the relationships were reassessed poststratification by gender, the strength of relationships between absolute values of maximum strength and performance substantially decreased and became nonsignificant for the smaller male group (n = 19). On scaling maximum strength values relative to BM, the strength of relationships increased for men and women with significant values obtained for each gender. The results of the study suggest when large variations in performance and strength exist within a sample, absolute values of maximum strength are sufficient to demonstrate relationships between the variables. In contrast, when variation is decreased, the magnitude of the correlation statistic is reduced (¹⁹), and normalization of maximum strength scores may be required to reveal relationships.

The majority of previous studies that have normalized maximum strength scores when investigating the relationship between strength and performance have performed so using simple ratio scaling (^2,23,30). This approach, however, assumes that there is a linear relationship between BM and strength. Most data contradict this assumption and demonstrate that while a positive relationship between strength and BM exists, the strength values of progressively heavier individuals fall below projected linear values (²²). More appropriate scaling may be achieved when using methods that assume nonlinearity between BM and strength. The theory of geometric similarity proposes that human bodies possess the same shape and therefore differ only in size (²²). As a consequence of this theory, it is predicted that any area measurement is proportional to BM raised to the power 2/3 (e.g., area ∝ BM^2/3). As force production is proportional to muscle physiological cross-sectional area (²⁶), based on the theory of geometric similarity maximum strength scores should be divided by BM raised to the power 2/3 when attempting to control for the effects of mass. This procedure has been used extensively to scale strength with BM in powerlifting and Olympic weightlifting (^5,26) and is commonly referred as allometric scaling in the literature. In the sport of Rugby union, there is considerable variation in BM across the different playing positions. Previous research has shown that, on average, elite forwards are approximately 20% heavier than elite backs (¹⁵). In addition, heavier forwards tend to produce greater maximum strength values, whereas lighter backs tend to perform better in tests that require high velocities such as in jumping and sprinting (¹⁴). The strong interrelationships between BM, strength, and performance are likely to require appropriate scaling practices if the effects of BM are to be controlled. Research by Crewther et al. (⁸) investigated the effects of ratio and allometric scaling of data collected from elite male rugby union players. Athletic performance was assessed primarily through short sprints and performance in jumping tests. Allometric scaling was shown to be superior to ratio scaling, as evidenced by greater reductions in correlations between performance and BM after values were normalized (⁸). In addition, the authors derived their own scaling exponents for the specific group of athletes based on the data collected. The derived exponents for BM and strength were shown to be consistent with the theoretical value of 2/3. In the present study, before scaling 1RM values relative to BM only trivial to moderate, nonsignificant correlations with performance measures were found. Conversely, after allometric scaling using the theoretical exponent of 2/3, both squat and deadlift 1RM values exhibited strong and significant correlations with all performance measures assessed. The results demonstrate that for sports such as rugby union where this is considerable variation in BM, maximum strength values should be normalized to provide appropriate context when evaluating players.

Because of the high velocity and explosive nature of many sporting tasks, a number of studies have attempted to quantify the relationship between variables such as power and RFD with measures of performance (^11,28,33). Results have been more varied than those quantifying relationships between maximum strength and performance. Some studies have reported strong relationships (^28,33), whereas others have reported independence between the factors (²⁴). Inconsistencies may be because of a number of factors, including the exercise and loads used to collect variables, the procedures used to calculate actual values, and the scaling methods implemented (¹¹). Some researchers have proposed that maximum strength acts as a general base influencing an athlete's ability to express high values of other key mechanical variables such as power and RFD (^7,34). These proposals are based on cross-sectional research highlighting strong interrelationships between the variables (¹⁰), and longitudinal studies demonstrating improvements in power and RFD when performing resistance training interventions designed to enhance maximum strength (⁷). Because of these strong interrelationships and proposed links between sporting activities and mechanical variables reflecting explosive force production, it has been suggested that performance in various sporting tasks may be better explained by combining mechanical variables rather than selecting single factors in isolation (¹¹). The regression models developed in the present study support this hypothesis. However, it is also clear that performance in the tasks studied in this investigation was principally explained by normalized maximum strength measures, and the addition of other mechanical variables such as average force, power, velocity, and RFD contributed substantially less.

The models obtained in the present study also revealed that the best combinations of variables were to some extent influenced by the nature of the activity. For example, RFD featured only in models for performance in the 5-m sprint and 505 agility test. In both these activities, performance is largely determined by the athlete's ability to accelerate their own BM from an initial state of low velocity. Research has shown that greater propulsive forces are generated when accelerating in comparison to sprinting at faster velocities (⁹). The combination, therefore, of both large maximum strength and RFD values would provide effective transition from low to high velocities in a short space of time. Based on similar reasoning, it may be expected that models explaining vertical jump performance, which also requires athletes to transition from low to high velocities would also feature maximum strength and RFD values. Indeed, a recent investigation established that peak RFD values measured during vertical jumps correlated strongly with jump height in physically active men (²⁸). However, the vertical jump features a single discrete movement in which performance is determined by the velocity at take-off, which is approximated very closely by the peak velocity obtained during the movement (²⁵). In contrast, performances in the 5-m sprint and 505 agility test are dependent on a more complex series of movements, which progressively increase the velocity of the body. Because of the cause and effect relationship between take-off velocity and performance in the vertical jump, it is unsurprising that the regression models identified peak velocity as a primary factor, especially as the testing movements were outwardly similar to the performance action. For the 10- and 30-m sprints, the regression models highlighted the combination of strength and power values as the best variables to explain performance. As the distance of the sprint increases, velocity and therefore contact time with the ground decreases (⁹). Mechanical power may reflect an athlete's ability to generate substantial ground reaction forces over short time periods and their capacity to store and release mechanical energy (³⁶), all of which would be important in influencing performance in these sprints.

It was hypothesized that the best models would generally combine maximum strength, anthropometric, and biomechanical variables. However, none of the best factor models included anthropometric variables other than BM, which featured directly as an explanatory variable and indirectly as a normalizing factor for maximum strength, force, and power data. Significant moderate to strong correlations were obtained between anatomical girths and performance measures. However, these relationships seem to be confounders for the primary inverse association that exists between BM and performance. All significant correlations obtained between anthropometric data and performance measures became nonsignificant after controlling for the effects of BM in the correlation analysis. The results from this study suggest that basic anthropometric factors such as girths, lengths, and widths have minimal effect on an athlete's ability to perform common sporting tasks such as running, jumping, and changing direction. However, it is important to note that in sports such as rugby, anthropometric factors can impact on an athlete's ability to perform the actual sport and successfully undertake tasks that are influenced by their role in the team (^17,18). In addition, BM and body composition may have important clinical relevance and roles in influencing performance. It has been suggested that increased skinfold thickness may provide a protective role against the high number of violent collisions sustained by rugby props (¹⁷). In contrast, higher body fat levels and overall mass may increase the likelihood of players suffering overuse injuries and can increase thermoregulatory stress endured by players, particularly in warm climates (^17,18).

The use of 2 distinct resistance exercises (deadlift and jump squat) to collect biomechanical data revealed that the explanatory power of regression models is influenced by the testing movement. In general, the 2 sets of regression models produced similar best explanatory models for each of the performance tests (Table 6), demonstrating that while combining specific mechanical variables with maximum strength may have a relatively small effect on the explanatory power of the model, selection of the most appropriate variables may be consistent. The models incorporating biomechanical variables collected during the jump squat consistently explained more of the variation in performance than those including data collected during the deadlift. This finding is in agreement with results published from a recent study investigating the relationship between power values collected during a traditional squat or jump squat with sprinting performance (³¹). Recruiting a sample of well-trained sprinters, Requena et al. (³¹) found that stronger correlation coefficients were obtained for absolute and relative power values collected during the jump squat in comparison with the same variables collected during the traditional squat. The authors proposed that the ability to accelerate the resistance throughout the entire concentric phase of the jump squat more closely match the kinematics of sprinting and therefore explain the stronger correlation values obtained (³¹). The concept of kinematic similarity was most evident in regression models applied to the vertical jump in the present study, where substantially more of the variance was explained by models featuring velocity values collected during the closely related jump squat in comparison with the same variables collected during the deadlift.

It is important to note the context in which the results from the present study were obtained. All measurements were made 6 weeks into the athletes' preseason training after a deload microcycle. The training block comprised primarily a maximum strength stimulus but included a smaller amount of work aimed at developing muscular power and sprint speed. Previous research conducted by Nimphius et al. (²⁹) using the correlation approach has demonstrated that with well-trained athletes, it is possible for the relationships between strength, power, and performance related variables to change over relatively small periods of time. Therefore, it is also possible that the models presented here were influenced by the athletes immediate training and may change during the course of a periodized program. Further research is required not only to monitor longitudinal changes between 2 variables but to determine if more complex interplays between more than 2 variables can also change substantially over time and exposure to different training content.

Practical Applications

From the results of this study, it is clear that the relative maximum strength of an athlete is the basic quality that determines their ability to perform many fundamental sporting tasks. However, relative maximum strength in isolation only explained approximately 35–65% of the variation in performance of the selected jump, sprint, and change of direction tests. Greater understanding and explanatory power was obtained by combining normalized maximum strength values with biomechanical variables measured during performance of explosive resistance exercises. For certain tests such as the vertical jump and 30-m sprint, as much as 90% of the variation (as measured by the unadjusted coefficient of determination) in performance of a field-based sports team was explained by combining the most suitable strength and biomechanical variables. The results also indicate that different biomechanical variables may relate more closely with performance in certain sporting tasks compared with others. Therefore, it is recommended that coaches continue to acknowledge that the development of their athletes relative maximum strength is of paramount importance. In addition, the results of this study support the use of contemporary training practices that include periodized programs aimed at developing maximum strength and additional features of the force-velocity curve. The results also suggest that the general preoccupation with mechanical power as a general descriptor of all biomechanical variables representing explosive athletic performance may be overly simplistic. In certain activities, the ability to displace relatively light resistances at high velocities or develop high RFD values may be more important to success than high mechanical power outputs. Importantly, the data also highlight that when an athlete has reached optimum BM for their sport, further improvements in strength and power should be achieved while trying to limit changes in their mass, as it is the relative values of these variables and not absolute values that have been shown to have the greatest influence on performance. Future regression models may wish to include technique-related variables to further enhance understanding of performance in tasks such as the 5-m sprint and change of direction tests where unexplained variance was still relatively high. In addition, more advanced body composition modeling including proportionality and segmental masses may provide data that combine more effectively with the force- and velocity-related variables identified in the present study.

Acknowledgments

The results of the present study do not constitute endorsement by the authors or the National Strength and Conditioning Association. The authors have no funding or conflicts of interest to disclose.