Understanding the physical demands of soccer requires accurate and objective quantification of the players' match activities (8,17,24,33). It is well established that soccer is characterized by low- (e.g., standing and walking) and high-intensity (e.g., running and sprinting) activities (5,22,25). Along with the sport-specific activities (e.g., tackles, turns, headers, dribbles), locomotor activities constitute the total physical load a player experiences during a match (26). Recently, attempts have been made to quantify this load through, for example, heart rate measurements of the players, estimates of running distance and intensity by video-based time-motion analysis, and various visual analog scale questionnaires (32,33). In addition, an increasing number of studies in the past few years have investigated different physical aspects of a soccer match with the use of computer-assisted motion analysis (3,8,10,14,30). The use of these kinds of systems has advanced our understanding of position-specific work rate profiles and the physical requirements of soccer players (10,32).
However, there are some important difficulties that must be recognized when evaluating physiological profiles on the basis of 2-dimensional (2D) analysis, which characterizes most of the match-analyzing tools in elite soccer. Although a large number of studies present the physical and physiological demands of elite soccer players on the basis of distance covered by running at different intensities (3,8,14,19,28,31), the analyses of differences in running speed and distance do not take into account some essential elements of soccer that together appear several hundred times in each match, for example, kicking and passing the ball, tackling, jumping, and turning (2,21,28). Two-dimensional systems compute expended energy only when the players travel from one location to another, with sprinting normally classified as the most strenuous movement. This may cause underestimation of the total energy expenditure because several of the players' high-intensity movements are made without changes in location on the pitch (10). Many 2D systems also assume that the players only travel in a forward direction, and hence, do not provide detailed information about movement such as backward or sideways running (10).
Many of the typical soccer activities have therefore been neglected, as previous studies only examine activities measured by distance and speed variables. Quite a few soccer-specific movements can cause significant physical stress on the players, although the distance and speed are low. Until now, these activities were counted and classified, but an objective measure of the load from these activities is still not developed. Therefore, there might be a potential for underestimating the external loads during a soccer match when only looking at results from time-motion analysis (18). Triaxial accelerometers are highly responsive motion sensors that record acceleration of body movement in 3 dimensions. These systems have been found useful not only for quantifying physical activity in a variety of populations but also for quantifying the physical and physiological demands in Australian football (6,7) and basketball (28). Recently, Scott et al. (34) found a higher correlation with internal training load for accelerometer load compared with high-speed runs and very high-speed runs during training. Although matches most likely will have higher distance of high-speed runs compared with training, these results suggest that additional monitoring tools are necessary to fully understand the external load during training and match play. It is well accepted that the activity profile of elite soccer players is position dependent according to running at different intensities (14,15,30). In soccer, therefore, the players' tactical roles and available space on the pitch will not only influence the numbers and distance of high-intensity actions the players are involved in but also in what way different positions achieve total load in matches. Motion analysis alone may thus underestimate the player load because frequently discrete high-intensity actions in soccer do not include running at high velocity (16). Moreover, this underestimation might be different according to player positions.
Another important factor concerning player load is movements with accelerations and decelerations. Movements with accelerations are more energetically demanding than constant velocity running (12,30). Even at a low running speed, a high metabolic load is imposed on a soccer player when acceleration is elevated (12,29). Decelerations are just as common as accelerations in soccer (29) and will therefore also contribute significantly to the players' load during match play. Earlier quantifications of accelerations in soccer have demonstrated a threefold to eightfold greater number of accelerations than sprints (8,36), but the number of studies looking into acceleration profiles in soccer are rather scant. To the best of our knowledge, only 2 previous studies have quantified decelerations in elite soccer players (4,29).
As a result of the fact that many typical soccer activities previously have been neglected in the evaluation of the player load during matches, the aim of this study was to combine triaxial accelerometer and time-motion analysis to obtain new knowledge concerning player load during match play. In addition, we aimed to determine acceleration/deceleration profiles of elite soccer players, and acceleration/deceleration contribution to player load during match play.
Experimental Approach to the Problem
To evaluate new perspectives about player load, acceleration, and deceleration profiles, this study used a fully automatic sport tracking system based on RadioEye (ZXY SportTracking AS, Radionor Communications AS, Trondheim, Norway) technology to assess match performances of professional elite football players over 3 full seasons. Total player load, player load from accelerations and decelerations, and number of accelerations and decelerations were analyzed for each half of the match to identify, and investigate differences between, the various player positions. Every player movement was captured by small body-worn sensors continuously monitoring the players' actions (www.zxy.no/zxy_sportchip.html). Data were transferred by microwave radio channel to 10 RadioEye sensors (ZXY SportTracking AS, Radionor Communications AS) mounted on the team's home arena. Player movement was registered at 20 Hz. The data were stored in an SQL server database, and the technology is proved to be ISO/ETSI/IEEE unlicensed ISM band standard (www.zxy.no/zxy_sensor.html).
The data set includes every domestic home game (n = 45) of over 3 full seasons (2009, 2010, and 2011) for the participating team (Rosenborg FC, age range: 25.38 ± 4.73 years). The matches were all played on grass surface. Movements of all players were observed and only data from players completing an entire match were used (n = 310, goalkeepers excluded). The sample included 8 central defenders (CD) (n = 68), 9 fullbacks (FB) (n = 83), 9 central midfielders (CM) (n = 70), 7 wide midfielders (WM) (n = 39), and 5 attackers (n = 50). Some of these players were used in different positions across, but not within, the matches included in the data material. Rosenborg FC won the Norwegian elite league and participated in the UEFA Europa League's qualifications during the first season of data collection. In 2010, the team again won the Norwegian championship and also participated in the Europa League's group stages. The third season (2011) ended with the team's bronze medal in the Norwegian elite league in addition to qualification matches for the Champions League and Europa League. Following an explanation of the procedures, all participants gave verbal and written informed consent to participate in the studies, and ethics approval was granted by the Ethics in Human Research Committee of the University.
Player load was based on acceleration measures from triaxial accelerometer sensors that were placed at each player's lumbar spine. The sensor was mounted in a specially designed belt wrapped tightly around the waist. The accelerometer registers data at 20 Hz and has a sensitivity of 184 μg/least significant bit (LSB), with a static noise of 1 mg. The signal was thereafter high-pass filtered (Butterworth first-order high-pass filter [frequency 20 Hz and frequency cutoff 0.0905]) to avoid the acceleration gravity (9.81 m·s−2) from being included in the final calculations.
Ideally, external workload is expressed in watts; in our case, the product of force and velocity or mass, acceleration, and velocity, with mass being a mere constant for each player, which can be ignored; the acceleration-velocity product provides a power value normalized for body weight. Obtaining accurate values for velocity from acceleration data is difficult mainly because of drift and uncertainty about initial conditions (starting velocity). Drift itself is small, but its accumulating effect in the integration process to velocity leads to large errors for velocity estimates. That is why we used acceleration data as a basis for load estimates, without integration to velocity.
Ultimately, the player load is calculated and presented as a downscaled (i.e., divided by 800) value of the square sum of the high-passed filtered accelerometer values for the respective axes (X, Y, and Z): ([X2] + [Y2] + [Z2])/800. Thus, the load value is the downscaled square of the player's absolute acceleration. The square was used to account for the effect of velocity that was not incorporated in the load algorithm. The downscaling was used for practical reasons.
Player load was also calculated per meter distance traveled (i.e., load per meter) and was examined for accelerations, decelerations, and over a full match.
According to the ZXY Sport Tracking system, accelerations and decelerations are defined by 4 event markers. First, the start of the acceleration/deceleration event is marked by the acceleration/deceleration reaching the minimum limit (1 m·s−2). Second, to be counted the acceleration/deceleration has to reach 2 m·s−2. Third, the acceleration/deceleration must remain above the 2 m·s−2 for at least half a second. The method for identifying accelerations and decelerations is not easy to contextualize within the available literature because there are so few studies and no standardized norm. Therefore, this study attempts to clearly state the method for identifying accelerations and decelerations in a way that could be replicated.
The following locomotor categories were selected: walking (from 0 to 7.1 km·h−1), jogging (from 7.2 to 14.3 km·h−1), running (from 14.4 to 19.7 km·h−1), high-speed running (from 19.8 to 25.2 km·h−1), and sprinting (≥25.2 km·h−1). The speed thresholds applied for each locomotor category are similar to those reported in previous studies (8,9,11,15,30). The above activities were later divided into 3 locomotor categories: total distance covered (sum of the distances covered during each type of activity), low- and moderate-intensity activities (locomotion <19.8 km·h−1), and high-intensity activities (locomotion ≥19.8 km·h−1).
Test-retest reliability of the sport tracking system is recently reported, indicating a good agreement for the x and y positions (from which, together with time, speed, and acceleration are derivatives) and the total distance covered (intraclass correlation coefficients [ICCs] were r = 1.0, 0.999, and 0.999 [p = 0.000], respectively) (23). Furthermore, there were no significant differences between the test and retest data for the same variables (Ibid). To assess the reliability of the testing systems calculation of player load, preliminary and unpublished data collected by our research group in a controlled on-field track indicate a test-retest reliability, assessed by the ICC, of r = 0.795 (p < 0.001). Also, Bland-Altman plots, as introduced by Altman and Bland (1983), indicated good agreement between the test and the retest data. We interpret these findings to indicate good reliability of the current tracking system (Figure 1).
The descriptive statistics were calculated and reported as mean ± SDs of the mean (SD) for each group of players on each variable. The differences between the independent variable player position groups in all measured variables (dependent variable) were examined using 1-way analysis of variance (ANOVA), followed by a subsequent Tukey post hoc test when differences were detected. A 1-way ANOVA was used to identify differences between each half, with paired t-tests to determine statistical significance. Statistical significance was set at p ≤ 0.05. To further assess the reliability of position and distance measures of the used tracking system, a 2-way, mixed intraclass correlation (ICC) reliability test between the measures taken while running rounds on a track was obtained, following the guidelines provided by Hopkins (20) for all measures. All data were then examined by a scatterplot, and a Pearson's r correlation coefficient was computed to determine the strength of the relationships between the paired variables from each round on the track. Also, a linear regression was drawn to the scatter plot and the shared variance (r2) and the equation of the predicted variable calculated. Furthermore, data were also plotted and investigated by using Bland and Altman's 95% limits of agreement, as described by Atkinson and Nevill (1). All statistical analyses were performed with SPSS 19.0 (SPSS, Inc., Chicago, IL, USA).
Over a full match, CD, CM, WM, and attackers had a 12, 18, 26, and 8% higher player load, respectively, than FB (Table 1). Also, WM had 13 and 17% higher player loads compared with CD and attackers, respectively, and their load trend was higher than CM (7%, p < 0.09). Central midfielders had a higher player load than attackers (9%). All playing positions had an approximately 5% decrease in player load from the first to second half. Findings from the first and second halves indicate similar patterns as found during a full match. The exceptions were that CM a had higher load compared with attackers during second half (9%, p = 0.05), that WM had a higher load than CM during first half, and that attackers only had a tendency toward a higher load than FB during second half (8%, p < 0.11).
Over a whole match, FB and WM accelerated more often than CD (39 and 43%), CM (15 and 18%), and attackers (15 and 18%) (Table 2). During the first half, FB accelerated more often compared with CD (30%), CM (19%), and attackers (19%), whereas WM accelerated more than CD (33%), CM (22%), and attackers (22%). During the second half, CD accelerated less compared with FB (46%), CM (36%), WM (50%), and attackers (36%). Player load from accelerations accounted for 8, 8, 7, 10, and 9% of total load for CD, FB, CM, WM, and attackers, respectively (Tables 1 and 2).
Over a full match, CD and CM decelerated fewer times compared with FB (55 and 27%, respectively), WM (50 and 22%, respectively), and attackers (48 and 20%, respectively). Also, CM decelerated more often compared with CD (23%). During the first half of the matches, FB decelerated more often compared with all other player positions, although only significantly more than CD (45%) and CM (28%). Also, WM and attackers decelerated more than CD (36 and 32%, respectively). Within the second half, CD decelerated less than FB (72%), CM (29%), WM (66%), and attackers (72%). Also, FB, WM, and attackers decelerated more than CM, with 29, 25, and 29%, respectively. Player load from decelerations accounted for 5, 7, 5, 6, and 7% of total load for CD, FB, CM, WM, and attackers, respectively (Tables 1 and 2).
The players on average covered a total distance of 10,200 ± 785 m in low-intensity activities (walking, jogging, running) during a full match. In high-intensity activities, results show a major difference between positions. Fullbacks' and WM's high-intensity running distances were >230, >48, and >40% more than CD, CM, and attackers, respectively. No differences were detected between FB vs. WM and CM vs. attackers in terms of high-intensity running. The distance covered in high-intensity running by CD was shorter than any of the other playing positions (p < 0.000) Differences in mean distances covered (±SD) for different positional groups are shown in Table 1.
For accelerations and decelerations together, the player load per meter distance was 1.65 (m·s−4). Therefore, the mean effort per meter was 28% higher for accelerations and 65% higher for decelerations when compared with the mean player load per meter for the other match activities. Also, there were positional differences in player load per meter distance (Table 3).
A novel finding from this study was that accelerations contribute to 7–10%, and decelerations to 5–7%, of the total player load across all playing positions during match play. Therefore, to obtain a “true workload,” accelerations, decelerations (i.e., number, distance, effort), and distance covered at various running speeds must all be included when evaluating a player's total workload during a full soccer game.
A main finding is that combined data from triaxial accelerometer and time-motion analysis demonstrate that player load is accumulated in a variety of ways across the different playing positions. Our data show that the use of only speed and distance variables to access the physical demands of a soccer player may be limited. This is because high-intensity bouts in soccer, such as jumping, tackling, collisions, accelerations and decelerations (duration <0.5 seconds), passing, shooting, and unorthodox movements (sideways and backward running), may be classified in the low-speed locomotor category, although there will be a high physical strain on the player. In light of this, a training load based on time-motion analysis or heart rate–based measures may not be a stimulating load for the players. We find that only using time-motion analysis may underestimate or overestimate how players are exposed to physical strain. For instance, when we compared CD and FB, CD covered the shortest distance for all locomotor categories, except for walking, and had the least accelerations and decelerations, but the highest player load and player load per meter. This finding highlights the potential application of accelerometers to measure player load at low velocities that may be underestimated for certain positions (7,18). Therefore, accelerometers may be a complementary tool for measuring the load from activities misrepresented by time-motion analysis (i.e., high-intensity bouts classified as low-speed activities), which, as we know from previous studies, occur several hundred times in a match (10,21).
We found that 12–16% of the total player load accumulates from accelerations and decelerations, indicating that the load from accelerations and decelerations constitutes a considerable part of the total load for a player during match play. This study reveals that elite soccer FB and WM accelerate more often compared with players in the other positions per match, whereas CD and CM decelerate less vs. players in other playing positions. Moreover, the study's results illustrate that the load per meter for acceleration and deceleration is higher than the load per meter for the full match. This emphasizes findings from Osgnach et al. (29), who show that a higher load is imposed on soccer players when acceleration is elevated. Furthermore, this study reveals that the load per meter for the full match also includes many sport-specific high-intensity activities (i.e., all the other soccer activities that do not include acceleration and deceleration for at least half a second). Therefore, a 28% higher load per meter during acceleration and a 65% higher load per meter in deceleration would indicate that the effort during accelerations and decelerations constitutes a considerable part of the total load for a player during match play. Osgnach et al. (29) estimated the energy cost of maximal deceleration (<−3 m·s−2) to be >3.41 J·kg−1·m−1, whereas the energy cost of maximal acceleration (>3 m·s−2) was >17.28 J·kg−1·m−1. Although this study does not estimate the energy cost, the relative level between acceleration and deceleration is not in line with Osgnach et al. (29), as the load per meter from decelerations is higher than from accelerations. The reason for this distinction may lie in the fact that Osgnach et al. examined the metabolic load, whereas player load in this study is from the players' accelerations of body movements (from the accelerometer) in 3 dimensions. Eccentric contractions, that is, deceleration, are less demanding metabolically, but a lot of external load comes from deceleration, which will not be captured by recording metabolic load.
Present data on average indicate a slightly, but significant, higher number of accelerations and decelerations during the first compared with the second half (∼38 vs. ∼37 and ∼28 vs. ∼27, respectively), and a total number of ∼76 accelerations and ∼54 decelerations. Other studies have previously quantified acceleration profiles of elite soccer players (8,36). Bradley et al. found that internationals and elites had an average of 119 accelerations during matches, whereas Varley and Aughey (36) obtained a mean of 115 accelerations (from low velocity to maximal) in elite Australian soccer players. To the best of our knowledge, only 2 previous studies have quantified decelerations in elite soccer players (4,29). Osgnach et al. quantified the distance of decelerations in the Italian first division and found distance decelerations (above 2 m·s−2) to be around 100 m more than in our study, although accelerations and decelerations in our study had to last for at least half a second. Bloomfield et al. (4) found an average of 54 decelerations in Premier League matches, which corresponds to the number of decelerations in this study. However, different methods, sport tracking systems, and variation in classifying accelerations make it difficult to conclude on the reason for this difference (32).
The average total distance covered during a full match (∼11,046 m) was well within the previously reported range for top players (8,14,35). Of the total distance covered, the amount of high-intensity activity accounted for ∼7.6% in the present cohort. This is slightly lower compared with previously published studies, as they found high-intensity activities to account for between ∼9 and ∼12% (8,30). The observed sprinting distances were on average shorter than previously reported for elite domestic and international football players (8,27), but comparable with UEFA Cup and Champions League matches (13). Players in wide positions sprinted longer distances compared with more central playing players during a full match. It has been speculated previously that a lower number of sprints among the more central players could be due to a lack of space for reaching the sprinting velocity (36). Also, the playing style, with emphasis on the need for the wider players to participate in both defensive and offensive phases of the match, has been proposed as a possible reason behind the elevated number of sprints for these players (13).
Although previous research indicated that various technological measurement systems for football match locomotion show relatively similar relative distributions of the various activities (e.g., running and sprinting) (32), it should be noted that different measurement technologies could cause the discrepancy in absolute measurements between the present and other studies. Hence, caution should be taken when comparing different studies of football match activities. Also, it cannot be ruled out that different styles of play, match score, the quality of opposition, or the fact that this study relies on one team's home games could be the reason behind the discrepancy in values of the different activities compared with other previously published research (36).
As mentioned in the methods section, our measure for load is based on acceleration only, and thereby biased toward forceful speed changes rather than high-speed runs. Thus, load at high but constant velocity runs will likely be underestimated. Future studies should tackle this problem by either combining this method with time-motion analysis or by developing algorithms that accurately combine acceleration and velocity.
In summary, the use of triaxial accelerometer technology in combination with time-motion analysis might be useful in assessing the players' load in soccer match play. Time-motion analysis is a useful tool of looking at the physiological demands from high-speed activities, but accelerometers may supply information concerning player load from the many discrete actions of a soccer match, which may be classified as low-speed activity. This seems especially important when looking into the player load from CD, as in this and most other studies, their amount of high-speed running, sprinting, and accelerations is appreciably smaller compared with the other playing positions. This might cause an underestimation of their load during matches and therefore an underestimation of the training load. Indeed, this study reveals some new factors concerning player load during matches, and that many high intensity actions without change in location at the pitch, have a significant contribution to the players load during matches and training. To design soccer training that is specific enough to meet the physical demands for each position, the coaches need a clear view in what way different players and positions achieve load. As the players in different positions achieve player load in a variety of ways, the principle of specificity suggests that different positions require distinct emphases on certain physical components to achieve a stimulating load in relation to the each positions requirements of a soccer match.
The authors thank the players for their effort throughout the period. T. Dalen and I. Jørgen contributed equally and share first authorship.
1. Atkinson G, Nevill AM. Statistical methods for assessing measurement error (reliability) in variables relevant to sports medicine. Sports Med 26: 217–238, 1998.
2. Bangsbo J. The physiology of soccer–with special reference to intense intermittent exercise. Acta Physiol Scand Suppl 619: 1–155, 1994.
3. Bangsbo J, Mohr M, Krustrup P. Physical and metabolic demands of training and match-play in the elite football player. J Sports Sci 24: 665–674, 2006.
4. Bloomfield J, Polman R, O'Donoghue P. Deceleration movements performed during FA Premier League soccer matches. J Sports Sci Med 6(suppl 6), 2007.
5. Bloomfield J, Polman R, O'Donoghue P. Physical demands
of different positions in FA Premier League soccer. J Sports Sci Med 6: 63–70, 2007.
6. Boyd LJ, Ball K, Aughey RJ. The reliability of MinimaxX accelerometers for measuring physical activity in Australian football. Int J Sports Physiol Perform 6: 311–321, 2011.
7. Boyd LJ, Ball K, Aughey RJ. Quantifying external load in Australian football matches and training using accelerometers. Int J Sports Physiol Perform 8: 44–51, 2013.
8. Bradley PS, Di Mascio M, Peart D, Olsen P, Sheldon B. High-intensity activity profiles of elite soccer players at different performance levels. J Strength Cond Res 24: 2343–2351, 2010.
9. Bradley PS, Sheldon W, Wooster B, Olsen P, Boanas P, Krustrup P. High-intensity running in English FA Premier League soccer matches. J Sports Sci 27: 159–168, 2009.
10. Carling C, Bloomfield J, Nelsen L, Reilly T. The role of motion analysis in elite soccer: Contemporary performance measurement techniques and work rate data. Sports Med 38: 839–862, 2008.
11. Di Mascio M, Bradley PS. Evaluation of the most intense high-intensity running periods in English FA Premier League soccer matches. J Strength Cond Res 27: 909–915, 2013.
12. di Prampero PE, Fusi S, Sepulcri L, Morin JB, Belli A, Antonutto G. Sprint running: A new energetic approach. J Exp Biol 208: 2809–2816, 2005.
13. Di Salvo V, Baron R, GonzÁLez-Haro C, Gormasz C, Pigozzi F, Bachl N. Sprinting analysis of elite soccer players during European Champions League and UEFA Cup matches. J Sports Sci 28: 1489–1494, 2010.
14. Di Salvo V, Baron R, Tschan H, Calderon Montero FJ, Bachl N, Pigozzi F. Performance characteristics according to playing position in elite soccer. Int J Sports Med 28: 222–227, 2007.
15. Di Salvo V, Gregson W, Atkinson G, Tordoff P, Drust B. Analysis of high intensity activity in premier league soccer. Int J Sports Med 30: 205–212, 2009.
16. Drust B, Atkinson G, Reilly T. Future perspectives in the evaluation of the physiological demands of soccer. Sports Med 37: 783–805, 2007.
17. Dwyer DB, Gabbett TJ. Global positioning system data analysis: Velocity ranges and a new definition of sprinting for field sport athletes. J Strength Cond Res 26: 818–824, 2012.
18. Gaudino P, Iaia FM, Alberti G, Strudwick AJ, Atkinson G, Gregson W. Monitoring training in elite soccer players: Systematic bias between running speed and metabolic power data. Int J Sports Med 34: 963–968, 2013.
19. Gregson W, Drust B, Atkinson G, Salvo VD. Match-to-match variability of high-speed activities in premier league soccer. Int J Sports Med 31: 237–242, 2010.
20. Hopkins WG. Measures of reliability in sports medicine and science. Sports Medicine 30: 1–15, 2000.
21. Iaia FM, Rampinini E, Bangsbo J. High-intensity training in football. Int J Sports Physiol Perform 4: 291–306, 2009.
22. Ingebrigtsen J, Bendiksen M, Randers MB, Castagna C, Krustrup P, Holtermann A. Yo-Yo IR2 testing of elite and sub-elite soccer players: Performance, heart rate response and correlations to other interval tests. J Sports Sci 30: 1337–1345, 2012.
23. Ingebrigtsen J, Dalen T, Hjelde GH, Drust B, Wisløff U. Acceleration and sprint profiles of a professional elite football team in match play. Eur J Sports Sci 15: 101–110, 2015.
24. Johnston RJ, Watsford ML, Pine MJ, Spurrs RW, Murphy AJ, Pruyn EC. The validity and reliability of 5-Hz global positioning system units to measure team sport movement demands. J Strength Cond Res 26: 758–765, 2012.
25. Krustrup P, Andersson H, Mohr M, Randers MB, Jensen JM, Zebis M, Kirkendal D, Bangsbo J. Match activities and fatigue development of elite female soccer players at different levels of competition. Reilly T., Korkusuz F., eds. In: Science and Football VI. New York, NY: Taylor & Francis, 2009. pp. 205–211.
26. Mallo J, Navarro E. Physical load imposed on soccer players during small-sided training games. J Sports Med Phys Fitness 48: 166–171, 2008.
27. Mohr M, Krustrup P, Bangsbo J. Match performance of high-standard soccer players with special reference to development of fatigue. J Sports Sci 21: 519–528, 2003.
28. Montgomery PG, Pyne DB, Minahan CL. The physical and physiological demands of basketball training and competition. Int J Sports Physiol Perform 5: 75–86, 2010.
29. Osgnach C, Poser S, Bernardini R, Rinaldo R, di Prampero PE. Energy cost and metabolic power in elite soccer: A new match analysis approach. Med Sci Sports Exerc 42: 170–178, 2010.
30. Rampinini E, Coutts AJ, Castagna C, Sassi R, Impellizzeri FM. Variation in top level soccer match performance. Int J Sports Med 28: 1018–1024, 2007.
31. Rampinini E, Sassi A, Morelli A, Mazzoni S, Fanchini M, Coutts AJ. Repeated-sprint ability in professional and amateur soccer players. Appl Physiol Nutr Metab 34: 1048–1054, 2009.
32. Randers MB, Mujika I, Hewitt A, Santisteban J, Bischoff R, Solano R, Zubillaga A, Peltola E, Krustrup P, Mohr M. Application of four different football match analysis systems: A comparative study. J Sports Sci 28: 171–182, 2010.
33. Rebelo A, Brito J, Seabra A, Oliveira J, Drust B, Krustrup PA. New tool to measure training load in soccer training and match play. Int J Sports Med 33: 297–304, 2012.
34. Scott BR, Lockie RG, Knight TJ, Clark AC, De Jonge XAKJ. A comparison of methods to quantify the in-season training load of professional soccer players. Int J Sports Physiol Perform 8: 195–202, 2013.
35. Stølen T, Chamari K, Wisløff U. Physiology of soccer: An update. Sports Med 35: 501–536, 2005.
36. Varley MC, Aughey RJ. Acceleration profiles in elite Australian soccer. Int J Sports Med 34: 34–39, 2013.