Time-motion analysis has been used extensively to investigate the movement patterns and physiological demands of high-intensity intermittent field sports (1,7,15,19). Recently, the time-intensive challenges of video time-motion analysis have been circumvented by the development and the application of global positioning system (GPS) satellite technology to the sporting environment (2,4,13,20). In addition to having the capabilities to collect and process large volumes of data very quickly, GPS technology provides quantitative information on the position, displacement, velocity, and acceleration of field sport athletes that could not be previously obtained from video time-motion analysis.
To simplify and summarize time-motion analysis results, locomotor categories have been used that are defined as a range of velocities (2,5,6), a mean velocity (>3 different means reported) (1,10,11), or various subjective descriptions of locomotor activities (e.g., ‘walking,’ ‘jogging,’ ‘striding,’ and ‘sprinting’) (9,16,17). More recently, some authors have simplified the terminology by using low-, moderate-, high-, and very high–intensity running classifications (3,8,14). Presently, there are no consistent definitions for velocity ranges making comparisons between studies problematic. The use of mean values is ambiguous for values that lie between 2 means. It also appears that in the existing analysis software and publications, velocity ranges have been determined arbitrarily (12).
A second and equally important question in GPS data analysis is the definition of what constitutes a sprint effort. Although few GPS time-motion analysis studies have been published, anecdotally sport scientists from professional and elite field sports typically characterize a sprint effort as any movement above a threshold sprinting velocity (ranging between 6 and 7 m·s−1). Given that a large proportion of field sport sprints involve maximal efforts of a short duration (∼1–2 seconds) (9,16,17), a limitation of this definition is that it fails to capture these short-duration, effortful movements that start at low velocities but do not achieve the sprinting velocity threshold.
Therefore, the purpose of this study was twofold: First, we investigated the velocity distribution of elite field sport athletes from 5 sports to statistically generate velocity ranges that could be consistently used for these sports. Second, we describe the process undertaken to develop a new definition of a sprint from GPS data that take into account both high acceleration and high velocity efforts.
Experimental Approach to the Problem
Data for this study were collected by trained sport scientists from universities and sport institutes across Australia, for the purposes of performance analysis. Data were collected from men's soccer (a professional Australian A-League team), women's soccer (a senior state league team), men's and women's field hockey (senior state league teams), and a professional Australian Rules Football (AFL) team. All the athletes were in peak physical condition and free from injury at the time of data collection.
Twenty-five data sets representing 5 complete games from 5 individuals were analyzed for 5 sports (a total of 125 data sets). Only data representing the actual play were included in the analysis. Warm-up, warm-down, and breaks in the game were removed from the data set. All data were supplied anonymously so that there was no way of identifying the individual athletes involved. The data came from GPS receivers designed specifically for use in sport (minimaxX, Catapult Innovations, Australia & SPI Elite, GPSports, Australia). Data were collected at 1 Hz, downloaded by the software that accompanies the GPS receivers and velocity data (measured by Doppler shift), and were exported for analysis. Several authors have found the accuracy of GPS technology for measuring the movement of athletes to be very good, including Townshend (18) who reported that 90.8% of GPS velocity measurements were <0.1 m·s−1 from actual velocity, and the mean distance error was 1.08 ± 0.34 m. The Institutional Ethics Committee approved all the experimental procedures.
Customized software (LabView, National Instruments, West Hartford, CT, USA) was used to calculate the frequency distribution of velocity (0–7.0 m·s−1 in increments of 0.1 m·s−1) for all the individual data sets, and the average distribution of velocity was calculated for each sport using Excel (Microsoft, Albuquerque, NM, USA). Decelerations or negative velocities were not included in this analysis. Additional customized software (Fortran 90, IBM, Armonk, NY, USA) was then used to find the optimal fit for 4 normal distribution curves (Gaussian) to the averaged velocity distribution curve for each sport (Figure 1).
Acceleration data (1 Hz) were calculated from velocity data for each data set. For each of the recommended velocity ranges (see Recommendations below), a frequency distribution of acceleration was calculated for each data set. The variance of this distribution was calculated to determine the threshold acceleration above which the highest 5% of accelerations exist (Figure 2). This threshold acceleration can be calculated for each velocity range and then used to identify sprints. This calculation is not made for the sprint velocity range because this movement had already exceeded the recommended threshold sprint velocity. This new data analysis process was applied to a single data set from a male soccer player to illustrate the typical results of the analysis. The existing or traditional definition of a sprint was also applied using the recommended threshold velocity of 5.5 m·s−1 (see Recommendations below).
The average velocity distribution determined for each of the sports was quite similar (Figure 3). They all had a characteristic peak at a typical average walking velocity (1.2 [range 1.1–1.3] m·s−1 or ∼4.3 [range 4.0–4.6] km·h−1) indicating that a large proportion of time is spent walking during play in each of the 5 sports. The relative amount of time spent standing is similar for all sports except for men's field hockey, for which it was twice as high as the others. Australian Rules football players tended to spend the highest relative amount of time running and sprinting.
Analysis of the velocity distributions from each sport produced similar velocity ranges (Table 1). The differences in velocity ranges that did occur tended to be larger between men's and women's sports than those among different sports.
The highest 5% of accelerations in each of the recommended velocity ranges (Table 2) were determined from a single data set of a male soccer player. The analysis revealed that the threshold acceleration used to identify a sprint tends to increase with increasing velocity.
When the sprint acceleration thresholds from Table 3 were applied to the sample data set, the analysis revealed 125 short sprints of high acceleration (highest 5%). These sprints were overwhelmingly short accelerations of a 1-second duration and covered distances of up to 4.2 m. There were also an additional 6 sprints that exceeded the threshold acceleration of >2- and 3-second duration. These sprints covered up to 7 and 7.9 m, respectively, but would not have qualified as traditional sprints because they did not reach the threshold sprint velocity (5.6 m·s−1).
The traditional sprint definition was also applied to the sample data set revealing a total of 19 sprints that exceeded a velocity of 5.6 m·s−1 (Table 4). More than half of the sprints were 1 second in duration covering a distance of up to 6.6 m. There were 8 sprints that were 2–3 seconds in duration that covered up to 11.9 and 18.2 m and reached a peak velocity of 8.0 and 6.3 m·s−1, respectively.
The implementation of the new and traditional definitions of a sprint indicates that sprints are typically 1–2 seconds in duration and cover 1.8–13.2 m. Neither definition of a sprint identified any accelerations >3 seconds in duration. The average peak velocity achieved during the most common sprints was 1.8 (0.4) and 5.8 (0.7) m·s−1 for the new and traditional sprint definitions, respectively. The total sprint distance was approximately 243M and 174 m for the new and traditional sprint definitions, respectively.
The intent of this work was to address the lack of standardized protocols for the analysis of numeric time and motion data from field sport athletes. We would have recommended that velocity ranges and sprint definitions published previously be adopted as the standard, but no single protocol appears to have been widely endorsed, and all appear to have been developed subjectively. Furthermore, the need for standards is important because we anticipate an increase in the publication of time-motion data in the future as the enabling technology (GPS, radiofrequency transponders, video analysis software, etc.) becomes more widely used.
The method used to identify velocity ranges in the present work made the assumption that each type of locomotion (walk, jog, run, and sprint) has a typical range of velocities that can be described by a distribution curve. This assumption was based on the observation that there is a peak in the velocity distribution curve of all data sets we analyzed that coincide with the typical average walking velocity (∼1.2 m·s−1 or ∼4.3 km·h−1). This velocity peak appears to be Gaussian in shape and represents the large amount of time spent walking during play. This feature of the data reflects what early time-motion analysis publications that used video analysis revealed—field sport athletes spend a large proportion of match play walking (9,15,19). Based on this finding, we extended our initial assumption further to include the presence of velocity distribution curves that represent the other locomotor categories. The next step was to attempt to identify these distribution curves in the overall distribution curve using a process of optimized multiple curve fitting. The intersection of these component curves was nominated as the boundary of each velocity range.
There are several limitations of this method including the likelihood that the component distribution curves are not normal and are almost certainly skewed. The component curves do not fit the overall distribution curve well where they intersect with each other. The locomotor category of “standing” had to be included as a velocity range because it is a relatively important part of time-motion analysis but excluded from the curve fitting process because, by definition, it should not involve any velocity. It was excluded from the curve fitting process but was arbitrarily defined as any velocity between 0.0 and 0.1 m·s−1. The velocity 0.1 m·s−1 was included in the standing velocity range because the velocity distribution curves of the data sets used in the present work include a “standing” peak that includes this value (Figure 1). The likely cause of this “standing” peak is the noise inherent in the GPS signal meaning that even when a GPS receiver is stationary, it reports very small changes in position and therefore velocity.
Overall, the velocity ranges for the 5 sports we analyzed were quite similar. The largest difference occurred at the threshold sprinting velocity between men's soccer and either women's soccer or women's field hockey. There was also a 0.7-m·s−1 difference between the threshold jogging velocity between AFL and either forms of field hockey. We consider the differences in velocity ranges between the 5 sports we analyzed to be relatively small, but the use of an averaged set of velocity ranges for all sports may cause slight overestimation or underestimation in the relative amount of activity in each locomotor category. Although, using sport-specific velocity ranges removes the opportunity to make comparisons between sports, we recommend that the sport-specific velocity ranges be used in the future. Where comparisons between sports are required, the averaged set of velocity ranges can be used.
The velocity ranges produced by the present method are similar to that used by Burgess et al. (2) but quite different to that reported by Di Salvo et al. (6). When the median velocity values reported by previous authors are converted to ranges, our ranges are very similar to those reported by Impellizzeri et al. (10) but quite different from the others listed in Table 5. The threshold jogging velocities are similar across all reports, but threshold walking and running velocities vary widely. Most of the previous reports use a threshold sprinting velocity of 6.0 m·s−1 or more, with the exception of Burgess et al. (2).
Results from video time-motion analysis have demonstrated that the majority of sprint efforts performed by field sport athletes are extremely intense but often short duration (∼1–2 seconds) in nature (2,17). Impellizzeri et al. (10) recognized the existence of short- and long-duration sprint efforts by using 3 different threshold velocities (“Running sprint [26.7 km·h−1], sprints <2 seconds [17.8 km·h−1], sprints >2 seconds [25.1 km·h−1]”). The majority of previous time-motion analyses of field sport athletes have defined a sprint as any movement reaching or exceeding a threshold velocity of 5.0–6.7 m·s−1 (Table 5). Krustrup et al. (11) defined a sprint as any movement reaching the sprint velocity range >6.7 m·s−1 (1) and recorded an average of 26 (range 9–43) sprints per game in professional soccer players. Mohr et al. (12) used the same definition and reported 39 ± 2 (mean ± SD) sprints lasting 2.0 seconds or longer, in professional soccer players. Burgess et al. (2) defined sprints as reaching 6.7 m·s−1 and lasting for 2.0 seconds or longer and reported 58 (±35) sprints per game in professional AFL players.
We cannot find a rationale in the time-motion analysis literature that justifies the use of a sprint velocity threshold or a minimum duration required to qualify as a sprint. However, the sample rate of the technology being used to measure velocity should be considered when attempting to identify short-duration movements in the data. Nevertheless, an examination of the frequency and range of accelerations that occur in the 5 sports we analyzed indicates that there are many types of sprints that involve high accelerations. These movements do not necessarily reach the typical sprint threshold velocity or minimum duration employed in previous reports but is no less important. We sought a new definition of a sprint that is based on velocity and acceleration. Because it is possible to reach a high velocity without high acceleration and vice versa, we set minimum thresholds for acceleration so that the definition identified movements that required great effort by the athlete. The threshold we chose was somewhat arbitrary but is applied in an objective manner. The choice of 5% as the acceleration threshold is analogous to the 5% significance that is widely used in statistical analysis. The inclusion of high acceleration, low-velocity efforts adds significantly to the understanding of field sport physiology.
In this study, we present a sample data set from a professional Australian soccer player, which contained 17 sprints of ≥1 second that exceeded our recommended threshold velocity of 6.1 m·s−1 for men's soccer. Using this traditional definition of a sprint, short-duration, maximal effort accelerations that start at low velocities and do not achieve the sprinting velocity threshold will be missed. Our analysis of a sample data set demonstrates that the highest 5% of accelerations at any speed or for any duration occur more frequently than do sprints defined in the traditional manner (n = 125 vs. 17 sprints). Furthermore, these high accelerations account for approximately 40% more total distance than do sprints defined in the traditional manner (243m vs. 154 m). Given that these accelerations are near maximal and that traditionally defined sprints are not necessarily associated with high acceleration, the relative energy expenditure is likely considerably greater.
In conclusion, this study developed velocity ranges and a standardized broader definition of a sprint (one that includes both high acceleration and high velocity efforts), for use in motion analysis of field sport athletes. The inclusion of high acceleration, low-velocity efforts adds significantly to the understanding of field sport physiology. In addition, the inclusion of high-intensity, short-duration sprint efforts provides conditioning coaches with specific information to train acceleration for field sport athletes.
We recommend that the sport-specific velocity ranges in Table 1 be used for time-motion analysis of field sport athletes. We also recommend that a sprint be defined as any player movement that satisfies one or both of the following criteria: (a) The movement reaches or exceeds the sport-specific sprint threshold velocity for at least 1 second and (b) the acceleration of the movement occurs in the highest 5% of accelerations in the associated velocity range. When the movement satisfies both criteria, it is counted as only 1 sprint.
Most studies define a sprint using any running that exceeds a threshold movement velocity; however, this definition fails to take into account short duration, maximal effort accelerations that do not reach high velocities. Given that the majority of sprint efforts performed in field sports are of a short duration, and involve maximal accelerations (2,16,17), valuable information could be lost if coaches only monitor high-velocity sprint efforts. We present specific velocity ranges to be used as definitions for locomotor categories and a new definition of a sprint that includes both high acceleration and high-velocity sprint efforts. The inclusion of high-intensity, short-duration sprint efforts provide conditioning coaches and applied sport scientists with specific information on the sprinting demands of field sport competition and a method of capturing the short-duration, effortful sprints performed by field sport athletes.
No financial support was provided for this research.
1. Bangsbo, J, Norregaard, L, and Thorsoe, F. Activity profile of competition soccer. Can J Sports Sci
16: 110–116, 1991.
2. Burgess, DJ, Naughton, G, and Norton, KI. Profile of movement demands of national football players in Australia. J Sci Med Sport
9: 334–341, 2006.
3. Coutts, AJ and Duffield, R. Validity and reliability of GPS devices for measuring movement demands of team sports
. J Sci Med Sport
13: 133–135, 2010.
4. Coutts, AJ, Quinn, J, Hocking, J, Castagna, C, Rampinini, E. Match running
performance in elite Australian Rules Football. J Sci Med Sport
, 13: 543–548, 2010.
5. Cunniffe, B, Proctor, W, Baker, JS, and Davies, B. An evaluation of the physiological demands of elite rugby union using global positioning system tracking software. J Strength Cond Res
23: 1195–1203, 2009.
6. Di Salvo, V, Baron, R, Tschan, H, Calderon-Montero, FJ, Bachl, N, and Pigozzi, F. Performance characteristics according to playing position in elite soccer. Int J Sports Med
28: 222–227, 2007.
7. Doğramaci, SN and Watsford, ML. A comparison of two different methods for time-motion analysis
in team sports
. Int J Perform Anal Sport
6: 73–83, 2006.
8. Duffield, R, Coutts, AJ, and Quinn, J. Core temperature responses and match running
performance during intermittent-sprint exercise competition in warm conditions. J Strength Cond Res
23: 1238–1244, 2009.
9. Gabbett, TJ and Mulvey, MJ. Time-motion analysis
of small-sided training games and competition in elite women soccer players. J Strength Cond Res
22: 543–552, 2008.
10. Impellizzeri, FM, Marcora, SM, Castagna, C, Reilly, T, Sassi, A, Iaia, FM, and Rampinini, E. Physiological and performance effects of generic versus specific aerobic training in soccer players. Int J Sports Med
27: 483–492, 2006.
11. Krustrup, P, Mohr, M, and Bangsbo, J. Activity profile and physiological demands of top-class soccer assistant refereeing in relation to training status. J Sports Sci
20: 861–871, 2002.
12. Mohr, M, Krustrup, P, and Bangsbo, J. Match performance of high-standard soccer players with special reference to development of fatigue. J Sports Sci
21: 519–528, 2003.
13. Pino, J, Martinez-Santos, R, Moreno, MI, Padilla, C. Automatic analysis of football games using GPS on real time. J Sports Sci Med
6 (Suppl 10): 9, 2007.
14. Rampinini, E, Impellizzeri, FM, Castagna, C, Coutts, AJ, and Wisloff, U. Technical performance during soccer matches of the Italian Serie A league: Effect of fatigue and competitive level. J Sci Med Sport
12: 227–233, 2009.
15. Reilly, T and Thomas, V. A motion analysis of work-rate in different positional roles in professional soccer match-play. J Hum Mov Stud
2: 87–97, 1976.
16. Spencer, M, Rechichi, C, Lawrence, S, Bishop, D, Dawson, B, and Goodman, C. Time-motion analysis
of elite hockey, with special reference to repeated-sprint activity. J Sports Sci
22: 843–850, 2004.
17. Spencer, M, Rechichi, C, Lawrence, S, Dawson, B, Bishop, D, and Goodman, C. Time-motion analysis
of elite field hockey during several games in succession: A tournament scenario. J Sci Med Sport
8: 382–391, 2005.
18. Townshend, AD, Worringham, CJ, and Stewart, IB. Assessment of speed and position during human locomotion using nondifferential GPS. Med Sci Sports Exerc
40: 124–132, 2008.
19. Withers, RT, Maricic, Z, Wasilewski, S, and Kelly, L. Match analysis of Australian professional soccer players. J Hum Mov Stud
8: 159–176, 1982.
20. Zubillaga, A, Gorospe, G, Hernandez Mendo, A, Villaseñor, AB. Analysis of high intensity activity in soccer highest level competition. J Sports Sci Med
6 (Suppl 10): 10–11, 2007.