Agility is defined as a rapid whole-body movement with changes of velocity or direction in response to a stimulus (13). This motor ability influences performance in sports where quick changes of direction are common (3,5). It is generally accepted that agility is a multifaceted quality that depends on several motor abilities, such as power, speed, and balance, and on morphological–anthropometric factors (11).
Because of the different agility manifestations (e.g., forward–backward, rotational, lateral, zig-zag, and “stop'n'go”), previous studies frequently investigated predictors of agility. The idea was to identify motor qualities that could be effectively trained and used to improve different agility capacities (9,15). However, nearly all the studies that evaluated agility predictors studied “nonreactive agility” (i.e., change of direction speed [CODS]). Agility in sport environments primarily depends on CODS and also on perceptual and cognitive processes. In short, perceptual and cognitive abilities allow an athlete to recognize the need to change direction and execute it effectively. In other words, in sports in which agility is an important quality, a change of direction is exclusively executed in response to unpredictable visual stimuli (e.g., opponent, teammate, and ball); therefore, agile maneuvers may not be explicitly preplanned (12,14).
Previous studies have already developed agility tests in which subjects have to quickly change direction in response to unpredictable visual stimuli (12,14). Regardless of the evident strengths of these methods (e.g., idea and development of the measuring procedures), we must emphasize certain limitations. In those studies, agility is observed through “nonstop running change of direction.” Basically, examinees sprinted over a “Y-shaped” course and changed their running directions only once. Because our respected colleagues were originally investigating rugby, where such maneuvers are meaningful and frequent, this was a logical experimental approach. However, there is no doubt that other sports demand more specialized reactive-agility tests.
In sports such as soccer, basketball, handball, and/or tennis, athletes have to change direction repeatedly throughout “stop'n'go” movements. While doing so, they often perform turns, alternate between running and lateral shuffling, change from forward to backward running, etc. The main difference between “stop'n'go” scenarios and the previously described “Y-shaped-course” scenario is that the latter scenario lacks a moment of “zero velocity” (i.e., “Y-shaped-course” agility consists of nonstop running). The distinction between “nonstop” and “stop'n'go” agility has been directly proven in recent studies, which demonstrated separate predictors for these 2 scenarios (10,11,15). Therefore, it is reasonable to conclude that the “Y-shaped course” is not an appropriate template with which to study reactive agility for all sports (12).
The first aim of this study was to evaluate the reliability and the validity of the newly constructed reactive-agility test to measure reactive-agility performance in sports in which “stop'n'go” agility exists. The secondary aim was to study the relationship between reactive and nonreactive agilities in college-level athletes.
Experimental Approach to the Problem
Previous studies have highlighted the importance of reactive agility. However, to the best of our knowledge, there is no study presenting testing protocols that would be useful in defining reactive agility in sports that involve repeated reactive “stop'n'go” directional changes (i.e., soccer, basketball, handball, and tennis). In this study, both within-subject and between-subject experimental designs were used to determine the reliability and the validity of the newly constructed “stop'n'go” reactive-agility test (SNG-RAT).
The experimental approach consisted of several phases. In the first phase, the test of nonreactive agility (i.e., CODS) and the complementary test of reactive agility were theoretically designed through consultations with renowned strength and conditioning experts from different sports (basketball, handball, and tennis), including consultations of professionals involved in the training of several national teams. After a consensus was reached about the test design, the approximate time frame and the most appropriate movement scenarios, the technical construction of the testing equipment commenced (i.e., the second phase of the investigation). We assembled and constructed digital equipment for the detection and recording of multiple time points throughout the tests. We constructed a novel hardware device system based on the ATMEL microcontroller (model AT89C51RE2; ATMEL Corp., San Jose, CA) as the core of the system. The photoelectric infrared sensor (E18-D80NK) was used as an external time-triggering input, and light-emitting-diode (LED) illuminations were used as controlled outputs. The photoelectric infrared sensor has been shown to be as reliable as high-speed sensors, with a response time of <2 milliseconds (>500 Hz) and a digital output signal. The detection distance of the sensor was from 3 to 80 cm, and was capable of detecting transparent or opaque objects. Because it has a digital output (high-low state) with an NPN transistor open collector, the sensor is connected through a microcontroller input-output port. For the purposes of our study, this device was connected to a laptop that worked on a Linux operating system (OS), although a MAC platform or other operating systems (e.g., Windows and Mac OS) would also be suitable. Measurement is more precisely explained later in the text. The third phase of the experiment included testing and reliability analysis for the newly constructed tests. The fourth phase involved validity analysis, through comparison of the characteristic groups of athletes (those involved in agility-saturated sports vs. noninvolved in agility-saturated sports).
College-aged athletes of both genders (N = 66), all 18–24 years old, were recruited for this study. The testing was performed as a part of the initial screening at the beginning of their respective competitive seasons. All subjects were in good health, based on an initial medical screening. In total, 3 participants had suffered recent musculoskeletal disorders (i.e., injury or prevalent pain) and were, therefore, not included in the study.
The participants were required to answer a questionnaire that was designed to assess the type of sport in which they were engaged. Additionally, their answers were checked by the university staff. If participants played in sports where running was not a regular part of the sports training and competition (e.g., rowing, swimming, synchronized swimming, water polo, and sailing), they were not included in this study (n = 6). As a result, a total of 36 college-aged male athletes (age 22.1 ± 2.4 years) and 21 college-aged female athletes (age 21.4 ± 2.5 years) participated. For the purposes of this study, the participants were additionally divided into 2 groups: those involved in agility-saturated sports ([AG]; soccer, basketball, handball, and volleyball; 11 women and 21 men) and those not involved in agility-saturated sports (nonagility sports [NAG]; gymnastics, track and field, and sport dance; 10 women and 15 men).
The subjects regularly participated in 8–14 hours of training per week, consisting of strength training (10–30%), endurance exercise (10–50%), and sport-specific exercise sessions (50–60%). All the measurement procedures and potential risks were verbally explained to each participant before obtaining their written informed consent. The Institutional Ethical Board gave written consent for the investigation after reviewing the experimental methods and procedures.
The measurements consisted of body height (BH); body mass (BM); body mass index (BMI), biceps, triceps, subscapular, and suprailiac skinfold thickness (for the calculation of the body fat percentage [BF%]); the SNG-RAT; the test of stop'n'go change of direction speed (SNG-CODS); and the countermovement jump (CMJ).
The BH and BM were assessed using a Seca stadiometer and weighing scales (Seca Instruments Ltd., Hamburg, Germany). The BMI was computed by calculating the ratio of BM (kilograms) and squared BH (meters). Body fat percentage was calculated using body density (BD) according to the following formula: BD = 1.162 − 0.063 × log Σ4KN (where Σ4KN = sum of biceps, triceps, subscapular, and suprailiac skinfolds). Body density was converted to body fat percentage: BF% = (4.95/BD − 4.5) × 100 (2,6).
The CMJ was used to compare the overall training status of the AG and NAG groups. The test began with the participant standing in an upright position. A fast downward movement to approximately 90° knee flexion was immediately followed by a quick, vertical movement as high as possible, all in 1 sequence. This test was performed without arm swinging and with the hands steadied by placing them on the hips.
The SNG-RAT was performed on the testing field as shown in Figure 1. The subjects began running from the start line when ready. Timing began the moment each subject crossed the infrared (IR) signal. When the subject crossed the IR signal, a hardware module (microcontroller—MC) ignited 1 of the 4 LED lights placed within 30-cm-high cones A–D. The subject had to assess which cone was lit, run to that particular cone, touch the top of the cone with his or her preferred hand, and return to the start line as quickly as possible. He or she had to then cross or step on the start line with his or her preferred leg, turn, and run again over the next course. Each time when the subject crossed the IR, the MC ignited one of the LED lights. The single-test trial consisted of 5 courses, and the single trial was completed when the examinee crossed the IR signal after returning from the fifth course. Three trials were performed. All the subjects were tested using 3 equal scenarios (i.e., 3 testing trials), though they had no knowledge of it in advance. The first scenario was 1-2-4-3-3, the second was 2-2-4-3-1, and the third was 4-1-4-1-2. The best result was retained as the final score. Although the testing equipment was designed to allow a random selection of testing scenarios, the same 3 scenarios were used to ensure equal testing conditions for all the subjects.
The SNG-CODS was performed on the same testing field as that for the SNG-RAT (Figure 1). Throughout this test, the testing scenario was simple (1,2), and the subjects knew it in advance. As for the SNG-RAT, the timing began the moment each subject crossed the IR. The subjects ran as quickly as possible to cone A (course 1), touched the top cone with his or her preferred hand and ran back to the start line. He or she then had to cross or step on the start line with his or her preferred leg, turn, and run over courses 2–5. The test was completed when the subject crossed the IR signal at the end of course 5. The test was repeated 3 times (i.e., 3 trials were performed) using the same scenario, and, after reliability analyses, the best score was retained as the final result.
As a measuring remark, it must be stressed that the MC recorded the result of each single course for both tests. On a day of testing, throughout the 2–3 practice trials before both tests (i.e., SNG-CODS and SNG-RAT), the subjects were familiarized with the testing procedures and established their most convenient maneuvers. More precisely, the examinees were instructed to use their preferred movement procedure and type of running (i.e., forward, backward, lateral displacement, and grapevine steps) and to strive for their best score.
On the first day of testing, data on the subjects' anthropometric measurements and CMJ were collected. On the second day, the SNG-RAT and SNG-CODS were evaluated. One half of the examinees conducted the SNG-RAT first, followed by the SNG-CODS, whereas the remaining half performed the SNG-CODS and then the SNG-RAT. Standardized 3-minute pauses between the trials and tests were used for all subjects. However, if the subjects asked for extra recovery time, this was allowed.
Each subject performed a standard warm-up that consisted of 5 minutes of forward running, 2 minutes of backward running, 1 minute of lateral shuffling (30 seconds each side), 10 squats, 10 push-ups, and 3 minutes of dynamic stretching exercises. Warm-up was performed before the CMJ testing on the first day and at the beginning of the second day of testing.
To avoid diurnal variation, all tests were carried out in the morning between 8 AM and 11 AM in December 2013.
The Kolmogorov–Smirnov test defined all variables as normally distributed. The reliability of the applied measurements was checked via their coefficients of variation (CV), Cronbach Alpha values (CA), and intraclass correlation coefficients (ICCs). The repeated measures analysis of variance (ANOVA), and Tukey's post hoc test were used to detect any systematic bias between the trials (items).
To determine the relationships between SNG-CODS and SNG-RAT, Pearson's correlations were calculated.
Differences between AG and NAG for all measured variables were established using an independent t-test. Additionally, the differences between AG and NAG were also analyzed using a magnitude-based Cohen's effect size (ES) statistic with modified qualitative descriptors. Effect sizes were assessed using the following criteria: <0.2 = trivial, 0.2–0.6 = small, >0.6–1.2 = moderate, >1.2–2.0 = large, and >2.0 very large differences (17).
All the coefficients were considered significant at a 95% confidence level (p ≤ 0.05). Statsoft's Statistic version 11.0 (Tulsa, OK, USA) was used.
In general, the reliability parameters suggest a high consistency for the SNG-RAT and SNG-CODS (ICC of 0.81–0.92). The between-subject reliability (CA of 0.79–0.91) and within-subject reliability (CV of 4–5%) were similar for both genders. There is no evident difference between the reliability of the SNG-RAT and SNG-CODS (Table 1).
The ANOVA found no significant differences between the SNG-RAT trials for women. Meanwhile, the ANOVA for repeated measures reached statistical significance (p ≤ 0.05) for the SNG-CODS (for both men and women) and the SNG-RAT (for men). As expected, there was a trend of improvement with repeated trials. In all cases where the ANOVA was significant, post hoc analyses revealed no significant differences between the second and third trials, demonstrating that the results were stabilized by the third testing trial. Consequently, the best score was retained as the final result for each subject on both tests (Table 1).
Correlations between the SNG-RAT and SNG-CODS were significant (p ≤ 0.05) but moderate (0.62 and 0.68 for men and women, respectively), demonstrating that reactive and nonreactive agilities share 36–46% of the common variance. Women and men performed better in the SNG-CODS than in the SNG-RAT (15 and 16% for women and men, respectively) (Table 1).
For men, the AG group achieved significantly better results in the SNG-RAT (t value = −2.28; p = 0.03; ES = −0.75, moderate differences), but no significant differences were observed between groups for the SNG-CODS (t value = −1.42; p = 0.17; ES = −0.47, small differences). No significant differences (p > 0.05) were observed between groups for anthropometrics (ES from 0.01 to 0.13, all trivial differences) and CMJ (t value = −0.93; p = 0.36; ES = 0.29, small differences), which indicated their similar overall training status (Table 2).
Among women, the AG and NAG did not differ significantly in SNG-CODS (t value = −0.30; p = 0.61; ES = −0.49, small differences) and SNG-RAT (t value = −0.88; p = 0.39; ES = −0.39, small differences). There were no significant differences for the anthropometrics (p > 0.05; ES from 0.04 to 0.40, trivial to small differences), but the ES for BF% indicated moderate (ES = 0.75) differences between groups (Table 3).
Regarding the main objectives of this study, there are several important findings. First, the newly constructed test was found to be appropriately reliable for both male and female college-aged athletes. Second, the moderate correlation between the SNG-RAT and SNG-CODS suggests that reactive agility and nonreactive agility should not be considered as single (i.e., equal) capacity. Third, there is evidence that the SNG-RAT superiorly measures the type of agility performance that is characteristic of agility-saturated sports.
The reliability of the different nonreactive-agility tests has been frequently studied. Because samples have differed between studies in terms of training status, gender, and sport, researchers have found moderate to high reliability parameters (4,7,8). In general, the reliability of the SNG-CODS is similar to previously reported values for other nonreactive-agility tests (3,4,9–11,15). This is particularly important because the SNG-CODS is relatively complex and consists of 5 courses and 5 unpredictable changes of direction. Each change of direction potentially causes some measurement error because of noncontrollable factors (e.g., incorrect stepping and sliding on the surface). Therefore, the high reliability of the SNG-CODS is more valuable.
The reliability parameters for the SNG-RAT suggest strong reliability in both genders. The ICC that was observed for men in this study was somewhat higher than the one previously reported for the Y-shaped reactive agility in rugby union players (ICC = 0.82) (12). However, this is almost certainly related to the higher level of heterogeneity in our subjects and the consequently higher variability of their results in comparison with equally trained rugby league players. To the best of our knowledge, no study has investigated reactive agility among women; therefore, the reliability of female reactive agility could not be compared.
Although all subjects were familiarized with the testing procedures, the ANOVA results indicated that testing on both procedures should be conducted for at least 2 trials. More precisely, the authors of the study witnessed 1 common mistake that was characteristic of the first trials. Initially, most of the subjects tried to sprint maximally from the start line to the infrared sensor (Figure 1). As a result, they accelerated uncontrollably, and their inertia did not allow them to perform the necessary change of direction efficiently. By the second or third trial, they anticipated this problem, which consequently led to superior and stable performances in the remaining trials.
A significant correlation between the SNG-RAT and SNG-CODS was expected because both tests were performed on the same course, and, as a result, the performance on both tests depended on similar conditioning capacities (i.e., running speed, power, and balance). Regardless of this significant correlation, however, it is evident that reactive and nonreactive agilities shared <50% of the common variance, which indicates that these 2 capacities should be observed as independent qualities (1,16,18). While studying rugby-specific agility, Australian authors identified an even smaller percentage of common variance between reactive and nonreactive components (12). However, the participants in that study watched a video recorded scenario before even starting the reactive-agility test (i.e., they stand still and received the video information), whereas the nonreactive test was performed over the same course with prior knowledge of the necessary movement pattern. Therefore, the majority of the reactive-agility test results in their study were associated with the early identification of key sport-specific kinematic cues from the video. On the other hand, our participants were already sprinting when they received visual information about the necessary change of direction. Therefore, we may suppose that perceptual and reactive capacities (P&RC, i.e., perception of the visual stimuli and reactions to stimuli) contribute less to the final result of the SNG-RAT than in the study that investigated the Y-shaped reactive-agility test. However, it is probable that the percentage of the common variance that remained unexplained after the SNG-RAT and SNG-CODS were correlated (∼60% unexplained variance) and can be attributed to the athletes' P&RC to some extent.
The only truly valid conditioning capacity test is one that efficiently differentiates the groups of interest and can consequently explain the difference between those groups (12). In our study, the validity of the test was demonstrated by its differentiation between 2 groups of athletes: those who were and those who were not involved in agility-saturated sports. Among the men, the SNG-RAT was found to be able to differentiate the AG and NAG athletes, whereas no significant differences were found for the SNG-CODS. Previous studies have confirmed that the nonreactive agility in men is mostly related to speed, power, and balance (10,11). Therefore, it is logical that those athletes who possessed advanced levels of these qualities would achieve good results in the SNG-CODS regardless of their P&RC. On the contrary, P&RC, especially visual stimuli, are factors that are systematically and nonsystematically trained almost exclusively in agility-saturated sports. Therefore, it seems reasonable to conclude that P&RC directly and indirectly influenced the reactive-agility performance of the AG group and defined this group's superiority in the SNG-RAT.
To some extent, the lack of significant difference between AG and NAG among women is contradictory to the previous discussion offered for men in which the SNG-RAT discriminated the 2 observed groups of subjects. However, this most likely does not mean that the SNG-RAT is not valid for women, but rather that certain modifications to the SNG-RAT test are necessary. For that purpose, we propose that shortening the test would be particularly useful. In short, there was certainly a possibility that other conditioning capacities (e.g., anaerobic lactate capacity) influenced the SNG-CODS and SNG-RAT performances among women. This was particularly probable observing the moderate differences in BF% between groups.
The hardware equipment designed for this study allowed us to measure and to record each testing sequence (i.e., the time necessary to complete each single course of the test). Therefore, we reanalyzed the test results for women, but this time as a final result, we have used the time necessary for them to complete the first 3 courses of the SNG-CODS and SNG-RAT (note that the original test result was recorded as the time necessary to complete 5 courses, which included 5 unpredictable changes of direction). After shortening the test, the results of the studied female groups were as follows: the AG group (women) achieved 6.71 ± 0.45 and 6.13 ± 0.56 seconds, whereas the NAG group's results were 7.55 ± 1.12 and 6.15 ± 0.87 seconds for the SNG-RAT and SNG-CODS, respectively. This time, the groups significantly differed only in the SNG-RAT (i.e., the AG group performed significantly better than the NAG group; t value = −1.98; p = 0.03; ES = −0.88).
This modification allowed us to determine the reactive-agility performance of women more accurately and consequently ensured the validity of the performance measurements. It is interesting to note that the between-subject-reliability parameters (CA) of both tests slightly improved when the shorter test version was applied among the women (0.83 and 0.90 for the SNG-CODS and SNG-RAT, respectively).
The originally developed reactive- and nonreactive-agility tests consisting of 5 courses were found to be appropriately reliable. However, the shorter version of the reactive-agility test (i.e., 3 courses) was able to differentiate the more agile female athletes from the less agile ones, and therefore, we strongly suggest the usage of a shorter version of the test (i.e., 3 courses) when testing female athletes. Although in this article, we presented a wired measurement system, the wireless version of the testing equipment is also in development. The authors are at disposal for all details about testing equipment availability, construction, and calibration.
We emphasize the high applicability of simultaneously performing both tests (i.e., reactive and nonreactive agilities). This implies that tests are performed over the same course and are therefore practically identical regarding the movement scenario. Therefore, the calculated ratio of the performance between the 2 tests (i.e., reactive-agility to nonreactive-agility ratio) would allow strength and conditioning coaches to indirectly determine the perceptual and reaction abilities (PRA) of the tested athletes.
The calculated PRA should be observed as a valuable conditioning parameter. It is reasonable to expect that PRA could be developed through the use of conditioning games or unplanned agility drills. This would also positively influence the agility that appears in the playing environment (i.e., decision-making agility).
Future studies should explore additional versions of the proposed tests and their applicability in sport-specific conditioning.
The authors are particularly grateful to the reviewers for their constructive comments and suggestions. Special thanks go to Mario Tomljanovic (Croatian Handball Federation), Mato Stankovic, Igor Jukic, and Mario Jelicic (Croatian Football Federation); and Stjepan Medak (Croatian Tennis Federation) for their valuable suggestions and overall help in this study. The results of this study do not constitute endorsement of the product by the authors or the National Strength and Conditioning Association.
1. Cavar M, Corluka M, Cerkez I, Culjak Z, Sekulic D. Are various forms of locomotion-speed diverse or unique performance
quality? J Hum Kinet 38: 53–61, 2013.
2. Durnin J, Womersley J. Body fat assessed from total body density and its estimation from skinfold thickness: Measurements on 481 men and women aged from 16 to 72 years. Br J Nutr 32: 77–97, 1974.
3. Hachana Y, Chaabene H, Nabli MA, Attia A, Moualhi J, Farhat N, Elloumi M. Test–retest reliability, criterion-related validity, and minimal detectable change of the Illinois agility test in male team sport athletes. J Strength Cond Res 27: 2752–2759, 2013.
4. Haj-Sassi R, Dardouri W, Gharbi Z, Chaouachi A, Mansour H, Rabhi A, Mahfoudhi M-E. Reliability and validity of a new repeated agility test as a measure of anaerobic and explosive power. J Strength Cond Res 25: 472–480, 2011.
5. Jakovljevic ST, Karalejic MS, Pajic ZB, Macura MM, Erculj FF. Speed and agility of 12- and 14-year-old elite male basketball players. J Strength Cond Res 26: 2453–2459, 2012.
6. Jelicic M, Sekulic D, Marinovic M. Anthropometric characteristics of high level European junior basketball players. Coll Antropol 26: 69–76, 2002.
7. Mirkov D, Nedeljkovic A, Kukolj M, Ugarkovic D, Jaric S. Evaluation of the reliability of soccer-specific field tests. J Strength Cond Res 22: 1046–1050, 2008.
8. Munro AG, Herrington LC. Between-session reliability of four hop tests and the agility T-test. J Strength Cond Res 25: 1470–1477, 2011.
9. Salaj S, Markovic G. Specificity of jumping, sprinting, and quick change-of-direction motor abilities. J Strength Cond Res 25: 1249–1255, 2011.
10. Sekulic D, Spasic M, Esco MR. Predicting agility performance
with other performance
variables in pubescent boys: A multiple-regression approach. Perc Mot Skills 0: doi: 10.2466/25.10.PMS.118k16w4, 2014.
11. Sekulic D, Spasic M, Mirkov D, Cavar M, Sattler T. Gender-specific influences of balance, speed, and power on agility performance
. J Strength Cond Res 27: 802–811, 2013.
12. Serpell BG, Ford M, Young WB. The development of a new test of agility for rugby league. J Strength Cond Res 24: 3270–3277, 2010.
13. Sheppard JM, Young WB. Agility literature review: Classifications, training and testing. J Sports Sci 24: 919–932, 2006.
14. Sheppard JM, Young WB, Doyle TLA, Sheppard TA, Newton RU. An evaluation of a new test of reactive agility and its relationship to sprint speed and change of direction speed. J Sci Med Sport 9: 342–349, 2006.
15. Spasic M, Uljevic O, Coh M, Dzelalija M, Sekulic D. Predictors of agility performance
among early pubescent girls. Int J Perf Anal Sport 13: 480–499, 2013.
16. Thomas JR, Nelson JK. Research Methods in Physical Activity (4th ed.). Champaign, IL: Human Kinetics, 2011.
17. Uljevic O, Esco MR, Sekulic D. Reliability, validity and applicability of isolated and combined sport-specific tests of conditioning capacities in top-level junior water polo athletes. J Strength Cond Res doi: 10.1519/JSC.0000000000000308, 2013.
18. Young WB, McDowell MH, Scarlett BJ. Specificity of sprint and agility training methods. J Strength Cond Res 15: 315–319, 2001.