The global positioning system (GPS) is a navigational system originally developed by the American Department of Defense for military use, but has since been released to the public and commercialized. It consists of 27 satellites equipped with atomic clocks in orbit around the earth. These satellites continually send information to GPS receivers and, using these signals, the receivers calculate the distance to the satellite (25). This is achieved by comparing the difference in time between the satellite's atomic clock encoded in the signal to the internal clock of the receiver. By calculating the distance to at least 4 satellites, accurate information can be generated regarding the GPS receiver's latitude, longitude, and altitude (25). Commercial GPS receivers, such as those used in team sports, also calculate speed using Doppler shift. This is attained by examining the frequency of the satellite signal, which is subject to change because of the movement of the receiver (25).
However, GPS have been known to have both technological and practical limitations. Satellite signals can be obstructed by the atmosphere and local environmental objects (e.g., stadiums and tall buildings) which can lead to measurement error (25). Likewise, the number of satellites interacting with the GPS receiver seems to play an important role in determining the accuracy of position estimates (29). Although 4 satellites are the theoretical minimum number needed to triangulate a GPS receiver's position, there is a moderate negative correlation between the total distance error (i.e., the difference between the recorded distance and the actual distance) recorded by a GPS receiver and the number of satellites signaling the receiver (16). Similarly, velocity measurements can be susceptible to a similar increase in error as the number of satellites interacting with the receiver decreased (38). Furthermore, the positioning of satellites interacting with the receiver is also vital in determining the accuracy of measurements recorded by the receiver (38). Quantification of the satellites distribution across the horizon is presented as the measurement known as the dilution of precision (DOP). This measurement is inversely proportion to the volume of a cone delineated by the position of the satellites in relation to the receiver (38). A DOP figure of 1 would represent the ideal distribution of satellites in the sky, for which 1 satellite would be positioned directly above the receiver with the remaining satellites equally distributed across the horizon. Contrastingly, if all satellites were positioned in a tight cluster directly above the receiver, the DOP would approach the maximum value of 50 (38).
In comparison with other tracking techniques, GPS is time efficient and allows for real-time feedback, allowing greater practicality in team sports. Indeed, where video motion analysis only has the ability to track one athlete's movements GPS units allow for simultaneous tracking of multiple athletes (2). Additionally, video analysis of player movement can take upward of 8 hours to extract variables reported in real time when using GPS devices (30). Furthermore, when attempting to estimate distance using video analysis, some methods require previous determination of mean speeds of each gait movement pattern as an attempt to overcome parallax error (10). When using GPS devices, manufacturers suggest that the devices be placed in an open area 15 minutes before use to allow the devices to lock onto satellites (11). Excluding this, there is no previous testing is required for use, increasing the ease and widespread usability of GPS devices.
With recent technological advances of GPS units and the development of other testing devices, athlete monitoring has seemingly become a more precise science. Athlete tracking using GPS was first actualized in 1997 and since then its use has become increasingly widespread and is currently commonly used in team sports, such as rugby league, rugby union, soccer, field hockey, and Australian football (9). Global position systems have the ability to objectively quantify the external training load of individual athletes during matches and training (28). Theory suggests that it is important for strength and conditioning coaches and sport scientists to understand the interplay of an athlete's internal and external training loads to monitor fatigue and optimize performance (3,19). The information gained from GPS devices during team sport training may allow for greater analysis and planning of periodized external training loads. However, subsequent match-play data may provide a greater understanding of the specific demands athletes face during matches. Allowing for tailored and match-specific training and recovery programs that may lead to increased performance and reduced injury prevalence (26).
Global positions systems are classified by the rate at which they sample per second. Initially commercial devices had a sample rate of 1Hz (one sample per second). There has been rapid advancement in the sampling rates of GPS and now 5, 10, and 15Hz units exist. Along with increases in sampling speed, the technological advancement of GPS has also seen the integration of triaxial accelerometers into the devices. Triaxial accelerometers use the sum of accelerations in 3 planes (X, Y, and Z) to produce composite vector magnitude (expressed as G-force). Accelerometers can quantify player/body load (the collation of all forces acting on an athlete) and impact measures by summing or reporting data on player contact and collisions with other athletes, objects, and surfaces (9).
To accurately interpret data, we must first understand the limitations of GPS devices. To do this, it is integral that the validity and reliability of GPS devices are examined. Validity and reliability are essential in scientific examination in sport as they allow greater interpretation and “meaningfulness” to results and findings. It becomes even more important when the obtained data are being used to prescribe, monitor, or alter an athlete's training regimes, as is the case with GPS use in many professional sport teams. This literature review will aim to collate all studies that have tested the validity and/or the reliability of GPS devices, at 1, 5, 10, or 15Hz, and accelerometers during simulations of either general team sport movements or movements that are specific to an identified team sport. The review will deliver a particular focus on the validity and reliability of GPS devices of all sampling speeds in relation to distance, speed/velocity and accelerations, and in the case of accelerometers player/body load and impacts.
To examine these aims, a comprehensive search of the online library of the University of Queensland, Google Scholar, and NCIB PubMed was performed to identify potential articles. The search strategy combined terms covering the topics of device (GPS, accelerometer), the sampling rate (1, 5, 10, and 15Hz), and the words validity and reliability. Studies were included if they had tested either validity or reliability of GPS or accelerometers with regard to distance, velocity/speed, acceleration, player/body loads, or a combination of these. This review is specifically concerned with the use of GPS in a team sport environment, and therefore, studies were excluded if testing protocols were outside the realms of use in team sports. Likewise, 1 study that met selection criteria was excluded because of it being in a foreign language with no translation available. From theses search parameters, 22 articles were identified and used in this review.
Validity and Reliability
The validity of an instrument reflects the ability of that instrument to accurately measure what it intends to measure (27). In a real-world application, this is vital to all facets of use regarding GPS; accurate reporting to exercise prescription, all rely on precise measurements by GPS units. The measures for validity are presented as the standard estimate of error (SEE), standard error of measurement (SEM), coefficient of variation (CV), or the percentage of difference of the mean from criterion measures. Reliability refers to the reproducibility of values of a test on repeat occasions (17). The reliability of measures is seen as critically essential when discussing the use of the GPS device in a real-life setting. The ability for a measure to be consistent with others (interunit) is needed to allow for comparison both between players and between sessions. Likewise, it is important for a GPS device to provide accurate information consistently (intraunit), as this allows comparisons between sessions. Measures for reliability are presented as CV or typical error of measurement (TEM), whereas intraclass correlations (ICC) have been reported when available. Based on previous recommendations, reliability is rated as good (<5%), moderate (5–10%), or poor (>10%) (12).
Currently, there are no recommendations for the acceptable error for validity measures in this field. However, for the sake of consistency and congruency we will adapt the recommendations for reliability to measures of validity. Therefore, measures of validity will be rated as good (<5%), moderate (5–10%), or poor (>10%). Not only does this assist in simplicity of reporting measures but for most team-based field sports, errors of up to 10% would be seen as an acceptable trade-off for time efficiency and ease of use. Furthermore, this review involves comparisons of different measures of difference or error (such as CV, SEM, TEM, etc.) This is due to a lack of consistency of reporting measures in the literature and should be noted that although related, different measures have different levels of conservativeness. Therefore, although grouping the variables allows for comparisons between studies, individual interpretation should be conducted with an understanding of the measure being taken, and the subsequent acceptable error in the sport, the measure would be applied. The values, or range of values for validity (Table 1) and reliability (Table 2) are documented in the tables below, followed by a comprehensive analysis of the individual GPS variables grouped by sampling rate.
1Hz Global Positioning Systems Devices
Global positioning systems sampling at 1Hz were the original units commercially available and were the initial GPS units used in team sport. Consequently, the largest amount of research and information available regarding the validity of GPS units relates to these devices. Most findings across these studies endorse the application of 1Hz GPS devices to track player movement during team sports. Edgecomb and Norton (13) performed 59 trials of a 1Hz GPS device after marked circuits (varying in distance from 128 to 1,386 m) and found that the GPS measures of total distance differed significantly from criterion measures, however, the average error of the GPS over the 59 trials was 4.8%. In the years since this initial study, there have been similar findings that further promote using 1Hz GPS devices in a team sport setting. One Hz GPS units were found to be capable of providing accurate information on distance for straight-line walking (1.79 m·sec−1) and low-speed running (3.58 m·sec−1) over a moderate distance (≈50–60 m), with the SEE being 2.7% and 2.6%, respectively (31). Additionally, 1Hz GPS devices are effective at producing accurate curvilinear distances (600–8,800 m) at varying velocities. Velocities ranging from walking (0–2 m·sec−1) to striding (4–5 m·sec−1) were all found to have acceptable validity when compared with the criterion distance, with the SEE ranging between 0.5% and 2.1% for measures (30). Furthermore, Gray et al. (16) found 1Hz GPS units produce valid measures of distance during 200 m linear and 200 m curvilinear courses. They found measurement bias (mean error) ranged between 0.8 and 2.8% for walking, jogging, running and sprinting during the linear course, and −0.5 and −9.8% during the curvilinear course. However, reducing the linear distance of tests seems to impair the accuracy of the distance measurement recorded by 1Hz GPS units. Walking, jogging, striding, and sprinting across shorter straight-line distances (10, 20, 40 m, and the 20–40 m interval) all failed to have acceptable levels of validity. The only exclusion was the 40 m walk which had a SEE = 9.6%, a moderate, although approaching poor validity (20). This finding identifies a potentially significant limitation of using 1Hz GPS devices in team sports. Sprints in team sports generally occur at very high intensities over short distances, commonly less than 20 m (34), therefore, it may be likely that 1Hz GPS devices will not be able to accurately report distances covered during these high-speed efforts. This may lead to an important part of an athlete's performance or sporting demands being misrepresented in reports.
Although understanding the validity of distance measures during these fundamental movements is vital, it is rare that movement in a team sport setting will involve purely linear or curvilinear running. Consequently, creating a course that replicates either general team sport movements or specific team sport movements is highly important in determining the ecological validity of GPS devices (26,31). Portas et al. (31) examined 6 courses incorporating multidirectional movement and changes of direction. This study found that 1 Hz GPS devices had good validity in tracking walking distances (1.8–4.2% SEE) and had good to moderate validity in tracking running movements (2.4–6.8% SEE) across all 6 courses (31). Likewise, creating a running circuit that included several changes of directions and intensities, Coutts and Duffield (8) found that 3 different models of GPS devices sampling at 1Hz all had good levels of accuracy for distance measures, with all producing results within 5% of criterion measures. Contrastingly, Jennings et al. (20) found that only walking during a course requiring gradual changes of direction had an acceptable level of validity for distance measures (SEE = 9.1%), and jogging during a course requiring tight changes of direction also had an acceptable level of validity for distance measures (SEE = 9.0%). Although walking and jogging varied between courses, striding and sprinting failed to have acceptable validity for distance measures (10.4–12.7% SEE) during both courses (20). These findings indicate that some movement occurring at higher intensities or across shorter distances with changes of directions might worsen the validity of distance measures by 1 Hz GPS devices.
However, in the same study, Jennings et al. (20) also found that when following a circuit designed to create general movement patterns and intensities commonly seen in team sport, 1Hz GPS devices were able to report total distance measures with good validity (SEE = 3.6%). The authors commented that it is likely that, despite involving both straight line and change of direction running at varying velocities, the increased distance of the team sport circuit leads to the greater accuracy of the 1Hz GPS reports for total distance (20). Furthermore, based on footage from English Premier League soccer matches, Portas et al. (31) designed courses specific to soccer defenders, midfielders, and forwards and also a course designed for a 1-minute bout of high-intensity activity. They found that 1Hz GPS devices were valid for calculating distances covered at high intensities for all positions in Premier League soccer and the bout of high-intensity activity (1.3–3.0% SEE) (31). Similarly, a study designed to test the validity of 1Hz GPS devices during field hockey movements found no significant difference between the total distance measured by the GPS device and the actual distance of the course (26). The same study showed significant differences between criterion distance measures and 1Hz GPS device distance measures during T-shaped shuttles, straight-line shuttles, zigzag shuttles and straight-line sprint shuttles. Despite these significant differences, the mean difference in distances between the 1Hz GPS unit and the criterion measures for all the various shuttles was small (±0.2 m) relative to the total distance measures of the shuttles (between 8.5 and 52.3 m) (26). These findings suggest that during team sport-like movement, the limitations observed during simplified running tasks are overcome and GPS devices can report an athlete's total distance accurately during team sport matches.
Despite the relative importance of reliability in GPS devices, both the intraunit and interunit reliability of 1Hz GPS devices is seemingly not well understood because of limited research and conflicting results. Although scientifically both forms of reliability are crucial for testing and comparison of players or comparison of scientific work itself, in a real-world team sport environment it is often the intraunit reliability that is more important for testing purposes. Intraunit reliability refers to the consistency of measures made by a single device over multiple testing bouts or occasions (8). Good intraunit reliability allows for comparison of an individual player's game-to-game or session-to-session activity, provided they are wearing the same device. A practice which is common in professional team sports. Research has shown that intraunit reliability is good during simple linear and curvilinear running (13,16,30,31). Initially, Edgecomb and Norton (13) reported moderate intraunit reliability values (TEM = 5.5%) for distance measurements during marked circuits varying in distance between 128 and 1,386 m. These findings have been improved more recently with 1Hz GPS devices showing good intraunit reliability (CV < 5%) when measuring straight-line, curvilinear and nonlinear distance, during walking, jogging, running, and striding over longer distances (200–8,800 m) and walking and running over moderate distances (≈50–60 m) (16,30,31). Likewise, when using courses involving multidirectional changes, 1Hz GPS units were found to have good to moderate levels of intraunit reliability for walking and running movements (31).
However, individual 1Hz GPS devices seem to have poor consistency over shorter distances. During a study that involved running across shorter distances (10, 20, 40 m, and a 20–40 m interval), 1Hz GPS devices were shown to have poor intraunit reliability of distance measures when walking, jogging, striding, and sprinting, with the exclusion of the 40 m walk and the 40 m jog (CV = 7.0% and 9.4%, respectively). The reliability of measures tended to worsen as the speed of the movement increased and the distance of the movement decreased (20). Jennings et al. (20) also found that the intraunit distance reliability of 1Hz GPS devices was only acceptable when measuring jogging during 2 courses created to cause gradual and acute changes of direction, respectively. Walking, striding, and sprinting were all found to have unacceptable levels of intraunit reliability (CV > 10%) during these courses. Although the findings of Jennings et al. (20) are without a common direction, the results do suggest that shorter distances during short linear and change of direction running protocols impair the intraunit reliability of 1Hz GPS devices.
However, similar to the validity of distance measures from 1Hz GPS devices, the increased total distance of team sport simulated circuits improve the intraunit reliability of distance measures from a 1Hz GPS device. Studies have shown 1Hz GPS devices have good to moderate levels of intraunit reliability (CV < 10%) when quantifying total distances during courses designed to recreate general team sport movements and courses typical of positional demands of soccer matches (20,31). This evidence suggests that although there are limitations that need to be acknowledged during both short distance during linear and change of direction running, these restraints are overcome during team sport movements. Individual 1Hz GPS devices will be able to consistently report total distance measures during team sport matches and therefore are appropriate to use in a team sports environment.
Interunit reliability refers to the ability of multiple devices to produce the same measure for the same experimental protocol (22). There has been considerably less research performed regarding interunit reliability, possibly because of the greater applicable nature of intraunit reliability as previously stated. However, interunit reliability is incredibly important when comparing scientific research or comparing multiple players during a single match or training session and therefore holds the same relevance scientifically as intraunit reliability. The findings regarding interunit reliability have been largely positive. Gray et al. (16) found that 1Hz GPS devices have good interunit reliability for walking, jogging, running, and sprinting during linear running (1.46–3.38% CV) and good to moderate interunit reliability for curvilinear running (1.63–6.04% CV). Similarly, Coutts and Duffield (8) have reviewed the interunit reliability of distance measures, although their results were mixed. They reported good to moderate interunit reliability for 3 different types of GPS devices sampling at 1Hz GPS for lap distance and bout distance (4.5–7.2% CV) during a running circuit. However, contrasting to the previous findings, Coutts and Duffield (8) reported very poor interunit reliability during high-intensity running and very high-intensity running (11.2–32.4% CV and 11.5–30.4% CV, respectively).
Therefore 1Hz GPS devices are able to consistently report distance measures and total distances are both comparable on repeat occasions using a single device and between devices. However, there should be care taken when comparing reported distances that occurred at a high intensity over short distance, both within device and between devices. Although there has been no research presented investigating the interunit reliability during change of direction courses, intraunit reliability seems compromised by these movements. Therefore, comparing distances during running involving multiple short changes of directions both between athletes or repeated measures from an individual athlete should be performed so with care. Additionally, there have been multiple reports that when using a 1Hz GPS device, the noise (measured as total error [TE] or coefficient of variation [CV]) may exceed the signal (smallest worthwhile change [SWC]), particularly at high intensities. This means that 1Hz GPS devices may not be capable of identifying the smallest changes of practical importance (20,31).
In terms of team sports analysis, the validity of speed/velocity measurements is equally important to the validity distance measures. Understanding the velocity at which athletes cover distance is crucial in establishing conditioning programs for athletes. Furthermore, accurate reporting of velocities by GPS devices can help reveal the sprint profile of an athlete and provide information on the relative amount of high-intensity work (work exceeding a standardized or individualized threshold velocity as defined by the practitioner and input into the GPS analysis software). These measures are important as relative work is considered a good indicator of team sport match-play or session intensity, whereas sprint performance is associated with success in high-level team sport competition (4,9). Only 2 studies were identified that tested the validity of speed or velocity measurements generated by 1Hz GPS devices. Barbero-Alvarez et al. (4) concluded that GPS measures were valid for estimating peak speed over a 30-m sprint, with significant Pearson correlations (p < 0.001) between GPS peak speed and total sprint time and fastest sprint time as recorded by timing lights (r2 = −0.96 and r2 = −0.93, respectively). Furthermore, after a course designed to simulate movements typical to that seen in matches of hockey, mean speeds recorded by timing gates and the 1Hz GPS units showed a very strong Pearson correlation of r = 0.99, interpreted by the authors as indicating good validity (26). The same study also found GPS mean speed measures to be not significantly different from timing gate mean speed measures for a T-shaped shuttle, straight-line shuttle, or zigzag shuttle (26). These results provide evidence that 1Hz GPS devices are able to report mean velocities during various multidirectional shuttles, matches, and single sprint efforts 30 m and over. However, during a shorter (13 m) straight-line sprint shuttle, GPS measures of mean speed were found to be significantly different from timing gate mean speed measures (26). The mean speed of an effort can be a good marker of intensity, and both these studies provide positive findings for the ability of 1Hz GPS devices to estimate these mean speeds. However, there have been no studies currently that have reviewed the ability of 1Hz GPS devices to accurately report instantaneous velocities during team sports, an equally if not more important marker of an athlete's performance during a single effort or a match. Nevertheless, these 2 studies provide preliminary framework to suggest that 1Hz GPS devices are capable of instantaneous or peak speed/velocity estimation during various shuttles (both linear and those involving changes of direction) and team sport movements.
Reliable measures of speed are important when attempting to quantify work by breaking it into intensity or speed zones. However, there is very limited research in this area with only 2 studies having reported reliability of speed measures by 1Hz GPS units. Barbero-Alvarez et al. (4) found that 1Hz GPS devices have acceptable intraunit reliability when measuring peak speed (CV = 1.2%, ICC = 0.97) and summated maximal speed (CV = 1.7%, ICC = 0.93) during a 30-m sprint. Likewise, only 1 study to date has reported interunit reliability of speed measures. Coutts and Duffield (8) found during a course designed to elicit high-intensity intermittent exercise seen in team sports that the interunit reliability of speed measures was good to moderate across 3 devices (2.3–5.8% CV). These findings suggest that speed measures obtained by 1Hz GPS devices are comparable both between- and within-devices during team sport testing and matches. However, because of the limited amount of research it should still be seen as best practice that an athlete wears the same device whenever data collection is occurring if at all possible. This negates any error caused by discrepancies in interunit reliability when comparing a single player's running profile on multiple occasions.
5Hz Global Positioning Systems Devices
Similar to the 1Hz GPS units, there is considerable literature that assesses the validity of distance measures reported by 5Hz GPS devices because of the length of time these have been commercially available for athlete tracking. The evidence largely suggests that 5Hz GPS devices are able to accurately quantify player distances during team sports. Five Hertz GPS devices have good validity when reporting straight-line distance when walking (SEE = 3.1%) or low-speed running (SEE = 2.9%) over a moderate distance (≈50–60 m) (31). Five Hertz GPS devices also have good validity when measuring curvilinear distances (600–8,800 m) at varying velocities from walking to running (0.4–3.8% SEE) (30). Furthermore, 5Hz GPS devices are a valid tool for measuring multidirectional walking and running distance (23,31). These results suggest that 5Hz GPS devices are able to accurately report distances during different velocities seen in team sports during moderate to long distance movements.
However, there have been concerns over the accuracy of moderate distance measures during high and very high-speed running. Rampinini et al. (32) found during multiple bouts of a 70-m straight-line intermittent shuttle course that the validity of distance measures for high-speed running were moderate (CV = 7.5%) and worsened greatly during very high-speed running (CV = 23.2%). Despite this, measures for total distance during the bouts of running were shown to have good validity (CV = 2.8%) (32). Furthermore, Jennings et al. (20) found that during a course instigating gradual changes of direction only walking and jogging produce acceptable results for distance measures (SEE = 8.9% and SEE = 9.7%, respectively). Comparably, when following a course that required tight changes of direction only walking was found to have acceptable levels of validity for distance measures (SEE = 9.9%) (20). In addition, Jennings et al. (20) also found when testing the straight-line distances of 10, 20, 40 m and the 20–40 m interval, only the 40-m walk and the 20–40 m interval for striding had acceptable levels of validity (SEE = 9.6% and SEE = 9.0%, respectively). Although these findings suggest there are similar limitations of high-velocity short distance running as seen in 1Hz GPS units, other research findings allude to a more complicated picture about the accuracy of 5Hz GPS units during these efforts.
All research to date into the validity of 5Hz GPS devices has used either the MinimaxX team 2.5 GPS device (5Hz, Catapult, Melbourne, Australia) or the SPI-Pro GPS device (5Hz, GPSports, Canberra, Australia). Although these devices have the same sampling speed, there seems to be a large discrepancy between the 2 devices ability to measure short linear high-speed running. Jennings et al. (20) investigated the validity of a MinimaxX team 2.5 GPS device, finding that this device is unable to produce results of even moderate validity (SEE <10%) during running and sprinting across shorter distances (as previously mentioned). Likewise, Petersen et al. (30) testing protocol included 2 MinimaxX Team 2.5 units and found similar results to those presented previously. During testing, only 1 MinimaxX device produced a moderate result for validity of distance, during a run-a-three sprint (SEE = 5.3%). During the same testing procedure, the second MinimaxX device produced a poor validity measure (SEE = 12.7%). Despite this, both MinimaxX devices had poor validity for measures of distance for sprints across 20, 30, and 40 m (14.4–23.8% SEE) (30). However, during the same testing protocol, Petersen et al. (30) also reviewed 2 SPI-Pro GPS units. They found that the SPI-Pro units had good to moderate validity for distance measures during 30 m and 40 m sprints and the run-a-three sprints (2.9–7.7% SEE), whereas 1 of the 2 units was found to have moderate validity for 20 m sprints (SEE = 5.5%). Furthermore, the same devices were reported to have good to moderate levels of validity (4.81–8.06% CV) when measuring distances from 10, 20, 30 m during sprinting, and a sprint that began with a moving 10 m (37). Consequently, evidence investigating 5Hz GPS units suggests that the SPI-Pro GPS device has acceptable validity when measuring shorter distance linear running that is occurring at high velocities, although MinimaxX team 2.5 devices may not be able to accurately report distances during these efforts.
However, the previously presented research using MinimaxX team 2.5 5Hz GPS devices are 2 and 5 years old, respectively (20,30). More recently published research from Vickery et al. (36) suggests that MinimaxX may have since overcome earlier problems and can now accurately estimate distances. It is likely that issues with the validity of the distance reported from earlier devices have been rectified through software updates. Research has previously shown software updates can alter the reporting of movement variables (6). Taking this into consideration, using change of direction courses similar to Jennings et al. (20), Vickery et al. (36) found that 2 MinimaxX 5Hz GPS units were able to produce results for distance that were not significantly different from criterion measures during both gradual and tight change of direction courses when sprinting. The same study also found that during a run-a-three protocol, cricket fielding protocol, and a random 10-second bout field-based team sports movement the two 5Hz GPS devices were not significantly different from criterion measures (36). However, during a fast bowling protocol (a 15-m sprint involving accelerations) distance measures from both devices were significantly different from criterion measures (36). This may indicate that some limitations involving the decrement of distance validity when sprinting during change of direction courses may have been overcome since earlier research. Although, distances during sprinting that has accelerations occurring may still be misrepresented by MinimaxX 5Hz GPS devices.
There is large importance placed on how a GPS unit can accurately report distances that are indicative of sport movements rather than simple straight linear and curvilinear distance. Therefore, it is important to examine the validity of 5Hz GPS units on these courses. Jennings et al. (20) found that when following a team sport simulated circuit, 5Hz GPS units were able to produce distance measures with good validity (SEE = 3.8%). Similarly, Johnston et al. (23) found that there was no significant difference between measures recorded by a 5Hz GPS device and criterion values during a team sport simulated circuit. Additionally, 5Hz GPS units produced valid measurements of distances position-specific running of soccer defenders (SEE = 2.2%), midfielders (SEE = 1.5%), forwards (SEE = 1.5%), and a 1-minute bout of high-intensity activity (SEE = 1.5%) (31). Interestingly, in all 3 studies the MinimaxX Team 2.5 GPS device was used. It is probable that the increased total distance of team sport simulated circuits causes the improvement in the validity of distance measures from these GPS devices, overcoming limitations seen in short distance and high-velocity running. This is despite the team sport circuits being primarily composed of shuttles of various distances (both linear and change of direction) being performed at ranging velocities. Therefore, both MinimaxX Team 2.5 and the SPI-Pro GPS devices would be appropriate to use for player tracking during team sport matches and training.
Currently, there is no clear evidence regarding the intraunit reliability of 5Hz GPS units during linear running, with similar research designs finding conflicting results. It seems intraunit reliability of both models of 5Hz GPS units is good (CV < 5%) during straight-line or curvilinear walking, jogging, striding, and running over both moderate (≈50–60 m) and longer distances (600–8,800 m) (30,31). MinimaxX 5Hz GPS units have also been shown to have good to moderate intraunit reliability when measuring multidirectional movements during walking and running (3.71–6.72% CV) (31). Indeed, Jennings et al. (20) found using MinimaxX 5Hz GPS units had moderate intraunit reliability when jogging, striding, and sprinting was performed across a course that involved gradual changes of direction (7.9–10.0% CV) and a course that involved tight changes of direction (8.6–9.7% CV). However, walking during both courses produced distance measures with poor intraunit reliability (CV = 11.5% and CV 15.2%, respectively) (20). The same study also found walking, jogging, striding, and sprinting all produce poor intraunit reliability results for distance measures across 10 m and 20 m (15.6–39.5% CV) (20). However, all movements categories produced distance measures with moderate intraunit reliability during a 40-m sprint (6.6–9.2% CV), whereas striding and sprinting also produce distance measures with moderate intraunit reliability during the 20–40 m interval (CV = 8.0% and 9.8%, respectively) (20). Despite this, the intraunit reliability of 2 MinimaxX 5Hz GPS devices to quantify distance when sprinting between 10 m and 40 m and a run-a-three remained poor (13.6–30.0% CV), with the exclusion of one of the 5Hz GPS units measure of a run-a-3 sprint (CV = 5.3%) (30). Although SPI-Pro GPS units have acceptable levels (CV < 10%) of intraunit reliability when reporting sprinting distances between 10 m and 40 m (30,37). It is difficult to interpret this collection of studies, as similar protocols have produced vastly different results even when the same brand of 5Hz GPS device has been used. The evidence would again suggest that the SPI-Pro 5Hz device may be preferable to the MinimaxX 5Hz devices, in this case because it is able to report short linear distances covered at high speeds. However, there is considerable research reviewing the MinimaxX device in comparison with the GPSports device and because of the variability of results from the MinimaxX device, this should be investigated further.
Despite these conflicting results, 5Hz GPS devices seem to produce good intraunit reliability results during team sport simulated circuits and team sport specific courses. Jennings et al. (20) found MinimaxX 5Hz GPS devices had good levels of intraunit reliability when reporting measures of distance (CV = 3.6%) during a team sport simulated circuit. Likewise, using a course designed to simulate the movements indicative of soccer defenders, midfielders, and forwards during matches and a bout of high-intensity movement, MinimaxX 5Hz GPS units were found to have good intraunit reliability levels for all measures of distance (2.21–4.49% CV) (31). These studies validate the use of a single 5Hz GPS device for player tracking and suggest that limitations seen during linear and curvilinear running are overcome during team sport movements.
Interunit reliability of 5Hz GPS devices is limited, although current research presents contrasting results. Jennings et al. (21) found that the interunit reliability of distance measures by two 5Hz GPS units was poor during linear running of 10, 20, and 40 m and the 20–40 m interval. Across all testing protocols, walking, jogging, striding, and sprinting had distance measurements that were greater than 10% difference between devices (10.2–11.9% difference), with one exception being the walking 20–40 m interval which had a difference of 9.9% between devices (21). Furthermore, they found that only jogging during a tight change of direction course produces mean difference values of less than 10% (difference = 9.5%), whereas during a gradual change of direction course only striding produces acceptable interunit reliability results (difference = 9.7%). They also found similar results when comparing two 5Hz GPS devices for both total distance and high-intensity running distance during team sports circuits (11.1% and 11.6% percent difference, respectively) and field hockey match-play (10.3% and 10.3% percent difference, respectively) (21). Similarly, in testing by Vickery et al. (36) the interunit reliability of two 5Hz GPS units was poor for 2 change of direction courses, run-a-three, fast bowling and fielding cricket protocol, and a 10-second bout of field-based team sport movement (17.7–22.8% CV).
Once again, it seems that some of these discrepancies between devices are largely overcome once movement begins to imitate team sport movements (i.e., multidirectional movement occurring over larger total distances), as opposed to shorter distance shuttles. However, it would seem that this is only the case for overall and low-intensity running distances. Johnston et al. (23) reported that during a team sport simulated circuit, 5Hz GPS units produced good interunit reliability measures for overall distance (TEM = 2.0%, ICC = 0.69) and produce acceptable interunit reliability measures of distances for walking, jogging, and running (TEM = 7.5%, ICC = 0.96; TEM = 8.2%, ICC = 0.29; TEM = 5.6%, ICC = 0.96, respectively). The interunit reliability of distance measures decreased for high-speed running (TEM = 10.8%) and decreased dramatically for sprinting (TEM = 112.0%) (23).
Collectively these studies present conflicting results, however both suggest that comparisons between devices of very high-speed running would be inappropriate. Because of the inconsistent findings it must be mentioned that applying the same device on a particular player across all sessions would be considered best practice. Furthermore, we suggest that coaches and sport scientists aim to use as few interunit comparisons as possible, until further research has been conducted on these devices. Concerns have previously been raised over the ability of to detect the smallest change of practical importance. Although, only 2 studies report the SWC, both found that for all measurements of distance the TE was much greater than measures of the SWC (20,31), indicating interpreting data in some cases may be problematic (18).
The largest amount of research into speed or velocity measurement validity has been conducted using 5Hz GPS devices. Four studies were identified as having assessed the validity of 5Hz GPS units to measure speeds or velocities. Waldron et al. (37) suggest that 5Hz GPS units have a moderate level of validity when measuring velocities during straight-line running (CV < 10%). The validity improved as the distance measure increased, but the measures with the best validity were speed measures beginning with a moving 10m sprint. Additionally, work from Varley et al. (35) found that 5Hz GPS units were able to produce valid measurements of instantaneous velocity when the initial velocity was high (5–8 m·sec−1) during constant velocity running (CV = 3.6%). Likewise, the 5Hz GPS device produced acceptable validity for measures of instantaneous velocity during running that involving accelerations when the run had moderate (3–5 m·sec−1) or high initial velocities (CV = 9.5% and CV = 7.1%, respectively) (35). However, Varley et al. (35) further presented that 5Hz GPS units had particularly poor validity for measurements of instantaneous velocity when initial velocity of the sprint was low (1–3 m·sec−1) or decelerations were occurring. Furthermore, Vickery et al. (36) found that two 5Hz GPS devices significantly underestimated mean speeds during both a tight and a gradual change of direction course and a fast bowling protocol (a 15-m sprint involving accelerations), whereas one of the devices significantly underestimated mean speed during a random field-based team sport movements. However, they found that peak speed measures by the two 5Hz GPS devices did not differ to criterion measures during all protocol (36).
Combined these results are important as they suggest 5Hz GPS devices can accurately measure velocities during sprints occurring after moderate velocity motion and involving accelerations. However, for running that involves the slowing or cessation of motion and running from standing or slow moving starts they might not be able to accurately report velocity measurements. Moreover, 5Hz GPS units were found to be a valid measure for instantaneous speed during a 50-m sprint after a moving start, and a valid measure for peak speed during a team sport simulated circuit (23). Again it seems that limitations seen during linear testing are not present during team sport testing indicating that these devices may have good ecological validity despite performing poorly during linear testing. However, there are concerns of measures of mean speed from 5Hz GPS devices during both sprinting involving accelerations and running involving change of directions.
Conclusions about the intraunit reliability of speed or velocity measurements from a 5-Hz GPS unit are difficult, as only 1 study has reported this. The study endorsed the repeat comparability of a single 5Hz GPS unit, reporting good intraunit measures of reliability when measuring speeds measures over 10, 20, 30 m sprints, a moving 10 m sprint, and overall peak speed (0.78–2.06% CV) (37). Similarly, only 2 studies have reviewed the speed and velocity interunit reliability of 5Hz GPS units, although the findings are ambivalent. Varley et al. (35) suggest that 5Hz GPS units have moderate levels of interunit reliability when producing measures for instantaneous velocity during constant velocity running when that running occurred with a moderate or high initial velocity (CV = 6.7% and CV = 6.3%, respectively). Varley et al. (35) also suggested that 5Hz GPS units have moderate interunit reliability when producing instantaneous velocity measurements during running involving accelerations when initial velocity was moderate (CV = 9.5%). However, the interunit reliability for instantaneous velocity measures was poor (CV>10%) when constant velocity running had low initial velocity, when running involving accelerations had low or high initial velocities and when running involving decelerations was occurring (35). However, Vickery et al. (36) found that during 2 change of direction courses, run-a-three protocol, and fast bowling and fielding protocol and a 10-second random field-based sports protocol that both mean and peak speed measures had poor interunit reliability (19.8–33.4% CV and 14.2–31.5% CV, respectively) (36).
During a team sport simulated circuit, 5Hz GPS units were found to have moderate interunit reliability when measuring peak speed (CV = 7.5%, ICC = 0.52) (23). Taken together, these findings again suggest that limitations seen during both linear and change of direction running do not translate to team sport movements and that multiple 5Hz GPS devices can be compared with each other during team sports. Nonetheless, based on the findings of Varley et al. (35) and given the limited research it would be best practice for athletes wear the same 5Hz GPS device whenever possible to allow for accurate match-to-match or session-to-session comparisons. Similar to distance measures, it has been suggested that 5Hz GPS may not be able to measure small and important differences in data because of the noise (CV) being larger than the signal (SWC) (35).
10Hz Global Positioning Systems Devices
Compared with the 1Hz and 5Hz GPS devices, there has been very little reported literature that has investigated the validity of 10Hz GPS units presumably because of their relatively recent development. Despite the limited research, early validity findings of 10Hz GPS devices are largely positive. Ten Hz GPS units produce acceptable measures of distance for short sprints (15 m mean SEM = 10.9%, 30 m mean SEM = 5.1%) (7). Although the mean SEM is greater than 10% (classifying as poor validity) for 15 m sprints, of the 9 devices 8 had SEM values less than 6% (5.4–5.8% SEM), whereas the remaining device recorded a SEM of 9.6%. Likewise, during a short sprint (5 m) from a walking start (cricket fielding protocol), a run-a-three protocol and during a 15-m sprint involving accelerations (cricket fast bowling protocol), Vickery et al. (36) found that distance measures from a 10Hz GPS device were not significantly different from criterion measures. Furthermore, Rampinini et al. (32) found that during intermittent shuttle running over moderate distances (70 m), measures of total distance and high-speed running distance from a 10-Hz GPS had good accuracy (CV = 1.9% and CV = 4.7%, respectively). However, accuracy worsened and became poor during very high-speed running (CV = 10.5%) (32).
Vickery et al. (36) found that measures of distance recorded by a 10-Hz GPS unit did not differ significantly to criterion measures during a 40-m running course requiring 7 gradual changes of directions. However, distance measures from a 10-Hz GPS unit were significantly different from the criterion measure during a shorter running course involving several tight changes of direction and during a random 10-second field-based team sport protocol (36). Despite this, when using a team sport simulated circuit there was no significant difference between the criterion distance and the total distance reported by a 10-Hz GPS unit (22).
From evidence to date, it would seem that 10Hz GPS devices are able to quantify short to moderate distances (<60 m) with higher accuracy when compared with the 1 and 5Hz GPS devices. Although 1 study has found that the 10Hz GPS devices may provide significantly different results to criterion measure during a tight change of direction course, the majority of the current evidence suggests that the 10Hz GPS data can validly measure distances during linear running and team sport simulated circuits at varying speeds and across varying distances. Findings suggest that the increase in sampling rate to 10Hz for GPS delivers superior validity for measuring distance when compared with the 1 and 5Hz GPS devices and therefore can be seen as preferable in a team sport environment.
Only 2 studies have reviewed the reliability of 10Hz GPS units. From the research available 10Hz GPS units have good intraunit reliability (CV < 5%) for measures of distance during 15 and 30 m sprints. However, the measurements over 30 m were reported to have greater stability than the 15 m measurements (7). Castellano et al. (7) also reported that the 10Hz GPS devices had good levels of interunit reliability when measuring distances of 15 m and 30 m sprints with CV = 1.3% and CV = 0.7%, respectively. Although these results are promising for the ability for 10Hz GPS devices to reliably measure running occurred over short distances at high speeds, there have been conflicting interunit reliability results reported during a team sport simulated circuit. The 10Hz GPS interunit reliability was good for total distance covered (TEM = 1.3%, ICC = 0.51), low-speed running distance (TEM = 1.7%, ICC = 0.97), and high-speed running distance (TEM = 4.8%, ICC = 0.88). When reporting distance covered during very high-speed running, the interunit reliability decreases, TEM = 11.5% (22). Although this suggests comparisons between 10Hz GPS devices during high-speed running would be unadvisable, all other interunit reliability measures have been good. It is therefore suggested that practitioners should use caution when comparing and interpreting high-speed running between devices.
Despite this, the research presented to date demonstrates that 10Hz GPS devices are able to accurately track fast movements across short distances with good intraunit reliability. Therefore, it should be suggested that players consistently wear the same device when possible. However, to date there has been no research to report the intraunit reliability of distance measures during team sport simulated circuits or sport-specific circuits.
Current literature suggests that 10Hz GPS devices may have overcome some of the problems observed with 5Hz GPS devices when attempting to measure instantaneous velocities. Ten Hertz GPS devices have good to moderate validity for measures of instantaneous velocities during constant velocity running and running involving accelerations, regardless of initial velocity. As the initial velocity of the run increased, the validity of instantaneous velocity measures improved. However, as with the 5Hz GPS devices, 10Hz GPS units have poor validity when measuring instantaneous velocities when decelerations are occurring (35). Similarly, Akenhead et al. (1) found that instantaneous velocity measures were valid for runs when the acceleration occurring between 0–4 m·sec−2 (0.12–0.19 m·sec−1 SEE). The same authors suggested that during accelerations greater than 4 m·sec−2 the validity of instantaneous velocity measures may be compromised (SEE = 0.32 m·sec−1) (1). Likewise, Vickery et al. (36) found that peak speed measures were not significantly different from criterion measures during a 15-m sprint involving acceleration (cricket fast bowling protocol), a 5-m sprint after a walking start (cricket fielding protocol), a run-a-three protocol and a tight and gradual change of direction courses, and a random 10- second field-based sport movement protocol (36). Mean speed measures were not significantly different from criterion measures during the fielding, fast bowling and run-a-three protocols, all of which involved short distance high-intensity running. However, during 2 change of direction courses and a random 10-second bout of field-based sports movements (both based heavily on multidirectional running) measures of mean speed were significantly different from criterion measures (36).
Despite these conflicting findings, only Johnston et al. (22) have investigated peak speed measures during a team sport simulated circuit. They found that peak speed measures were significantly higher than criterion measures during the team sport simulated circuit for 2 GPS devices (p ≤ 0.01) (22). However, they reported Pearson's correlates (r) of r = 0.89 and r = 0.91, respectively (22). This finding reveals that 10Hz GPS devices may overestimate athlete peak speed during team sport matches, giving false information on max speeds reached during individual efforts. Coupled with research in linear and change of direction running, it would seem there may be some limitations around the ability of 10Hz GPS devices to measure both mean and peak speed during team sport running and running involving frequent change of direction. Furthermore, 10Hz GPS devices seem to poorly estimate instantaneous velocity when very high accelerations (>4 m·sec−2) are occurring during player tracking in team sports.
Although no studies have looked at the intraunit reliability of 10Hz GPS devices, three studies investigating the interunit reliability of 10Hz GPS units provide encouraging results during sprinting and team sports simulated movements. These studies demonstrate that 10Hz GPS units have good to moderate interunit reliability for measurements of instantaneous velocity during running involving accelerations (1.9–4.3% CV), constant-velocity running (2.0–5.3% CV), or during running incorporating decelerations (CV = 6.0). Furthermore, 10Hz GPS devices can produce reliable instantaneous velocities measurements regardless of the initial velocity of the run being performed (35). Indeed, when undertaking constant velocity running and running involving accelerations, as the initial velocity increased the interunit reliability improved. Additionally, 10Hz GPS units have been found to have good interunit reliability when measuring instantaneous velocity during a 10-m sprint regardless of the magnitude of the mean acceleration taking place (0.7–9.1% CV), although the reliability worsened as the magnitude of the acceleration increased (1). Last, Johnston et al. (22) found that 10Hz GPS devices had good interunit reliability when measuring peak speed during a team sport simulated circuit (TEM = 1.6%, ICC = 0.97).
Taken together these finding suggest that regardless of the movement taking place, velocity measures are able to be compared between separate 10Hz GPS devices with confidence. This also means that if athletes are unable to wear the same device over multiple tracking sessions that their data are still comparable. Likewise, it is possible to use a 10Hz GPS unit to compare velocity or speed measures between athletes during a single session or over separate sessions confidently. Importantly, there is evidence to suggest that 10Hz GPS may be able to quantify the smallest change in speed measures that would be deemed worthwhile. Varley et al. (35) found when using 10Hz GPS devices to measure instantaneous velocities during running that all noise values (CV) were much smaller than the signal values (SWC), indicating that 10Hz GPS devices are sensitive enough to detect all physiologically significant changes in velocities.
15Hz Global Positioning Systems Devices
The research on 15Hz GPS devices is incredibly limited based on the short amount of time since its development. To date, only 3 studies have reviewed the validity of distance measures from 15Hz GPS devices. Rawstorn et al. (33) found during a shuttle running protocol designed to recreate soccer movements that both linear and curvilinear distance was significantly different from criterion measures. However, the measurement bias (mean error) was good for walking, jogging, running, and sprinting for both the linear shuttle (2.95–3.16%) and the curvilinear shuttle (−2.20 to 1.92%) (33). Conversely, when using two 15Hz GPS devices, Vickery et al. (36) found that both devices differed significantly on distance measurements to criterion measures during a gradual change of direction course, although one was significantly different during a tight change of direction course. However, during a multidirectional 10-second bout of team sport movements both devices did not differ significantly to criterion measures (36). Furthermore, Vickery et al. (36) found that both devices were not significantly different from criterion measures during short high-intensity linear-based running (run-a-three, cricket bowling, and cricket fielding protocols). Likewise, Johnston et al. (22) found no significant difference between the criterion distance and the total distance reported during a team sport simulated circuit. Although this research area does need to be expanded, these initial findings are promising when considering the use of 15Hz GPS units during team sport use.
Fifteen Hertz GPS units have been shown to have good intraunit reliability during soccer specific movements. Rawstorn et al. (33) found that 15Hz GPS devices have good intraunit reliability during a soccer specific shuttle test when both linear shuttle running was occurring (CV = 2.44%, ICC = 0.99) and when curvilinear running was occurring (CV = 2.16%, ICC = 1.0). Although there has been seemingly little research performed regarding interunit reliability of 15Hz GPS units, the current evidence is somewhat contradictory. Vickery et al. (36) found that interunit reliability was moderate between 2 devices during a fast bowling protocol (15 m sprint involving acceleration), a gradual change of direction course and a 10-second bout of random field-based team sport movement (CV = 5.5%, ICC = 0.55, CV = 6.2%, ICC = 0.46, and CV = 8.2%, ICC = 0.10, respectively). However, interunit reliability was poor during a high-intensity run-a-three, cricket fielding protocol, and a tight change of direction course (12.5–17.9% CV) (36). Conversely, Buchheit et al. (6) found that 15Hz GPS units have good to moderate interunit reliability when reporting total distance, low-speed running (speed) and high-speed running (speed) (CV = 3%, CV = 2%, and CV = 6%, respectively). Johnston et al. (22) found similar results regarding in interunit reliability when reporting total distance (TEM = 1.9%, ICC = −0.20) and low speed running distance (TEM = 2.0%, ICC = 0.98). However, in contrast to the previous study, Johnston et al. (22) found when using a 15-Hz GPS unit to measure high-speed running distance, the interunit reliability decreases (TEM = 7.6%, ICC = 0.94) and becomes poor when measuring very high-speed running distance (TEM = 12.1%).
Interestingly, this previous study by Johnston et al. (22) reviewing the interunit reliability of 15Hz GPS devices was performed in conjunction with 10Hz GPS device testing and the interunit reliability for distance measures by the 15Hz GPS devices was poorer during all measures (22). This suggests that increasing the sampling rate from 10Hz and 15Hz with GPS devices may actually decrease the interunit reliability of distance measures. Although this conclusion is hard to determine given the limited research comparing these 2 devices. Findings from the combined studies suggest that distance measures from 15Hz GPS units are comparable within device but, because of a lack on conformity of evidence, during high-speed running comparisons between devices should be performed so with caution.
Vickery et al. (36) found that during short high-intensity running protocols (cricket fast bowling, cricket fielding, and a run-a-three protocol) that two 15Hz GPS devices had both peak speed and mean speed measures that did not differ from criterion measures. However, during a gradual change of direction course, both devices had mean speed measures that were significantly different from criterion measures, although both recorded peak speed measures that were not significantly different (36). During a tight change of direction course one device significantly underestimated mean speed, whereas the other device significantly overestimated peak speed (36). However, these findings were not translated to a heavily multidirectional 10-second bout of random field-based team sport movements (36). During a team sport simulated circuit, Johnston et al. (22) found one GPS device differed significantly (p = 0.009) from criterion peak speed measure, whereas the second device did not (p = 0.76). However, the Pearson's correlates (r) for the 2 GPS units were large (r = 0.64) and very large (r = 0.76), respectively. Again, although these measures are similar to the results reported for 10Hz GPS devices in the same investigation, the Pearson's correlations were stronger for the 10Hz GPS devices. This suggests that there may be no additional benefit to the increased sample rate that 15Hz GPS devices offer. In fact, the increase in sample rate may actually be detrimental to the accuracy of speed and velocity measurements. The overall research suggests that mean speed measurements from 15Hz GPS devices should be interpreted with caution.
Although only 1 study to date has reported the interunit reliability of mean speed measures, three studies have reviewed the interunit reliability of 15Hz GPS units when measuring peak speed. Mean speed has been reported as being moderate during a gradual change of direction course, a cricket fast bowling protocol (15 m sprint involving acceleration) and a random 10-second bout of field-based team sports movements (CV = 7.8%, ICC = −0.14, CV = 8.8%, ICC = 0.50, and CV = 7.5%, ICC = 0.39) (36). Interunit reliability for mean speed measures was poor for a course causing tight changes of direction, a run-a-three protocol, and a cricket fielding protocol (10.9–16.3% CV) (36). Using the same protocol, Vickery et al. (36) also found that peak speed measures during a cricket fast bowling protocol (15 m sprint involving acceleration) were moderate (CV = 8.4%, ICC = 0.36), although interunit reliability of peak speed measures was poor during tight and gradual change of direction courses, a cricket fielding and a run-a-three protocol, and a 10-second bout of random field-based team sports movement protocol (11.9–20.0% CV). These findings regarding mean speed measure directly oppose others using 15Hz GPS devices. During a standardized running routine, 15Hz GPS units were found to have good interunit reliability for measures of mean speed (CV = 1%) (6). However, during a team sport simulated circuit, 15Hz GPS units were reported as only having moderate interunit reliability (TEM = 8.1%, ICC = −0.14). Furthermore, the results were worse than those found by 10Hz GPS devices during the same testing protocol (22). The findings of validity and reliability of 15Hz GPS units to date suggest that there is no advantage to using these devices over a 10Hz GPS device.
Currently, only Buchheit et al. (6) have reviewed the reliability of acceleration measures by any GPS unit. Using 15Hz GPS units during a 30-minute standardized running routine, they found peak acceleration measures had poor interunit reliability (CV = 10%), whereas accelerations greater than 3 m·sec−2 and accelerations greater than 4 m·sec−2 had very poor interunit reliability (CV = 31% and CV = 43%, respectively). Furthermore, decelerations of the same magnitudes (>3 m·sec−2 and >4 m·sec−2) also had very poor interunit reliability (CV = 42% and CV = 56%, respectively) (6). The large range of interunit reliability suggests that comparisons of accelerations between devices should not be recommended. Furthermore, the article does not comment on the validity of measurements and therefore reporting and comparison of this variable should be performed so cautiously.
The integration of triaxial accelerometers into GPS devices has allowed for a greater scope of information to be recorded about an athletes work rate patterns and physical loads (9). As previously stated, the triaxial accelerometer summates accelerations in the 3 movement axes (X, Y, and Z planes) to measure a composite magnitude vector (expressed as a G-force) (9). This addition to the GPS device allows the total number and intensity of an athlete's collisions and contacts during a team sport match can be quantified by impact measures or body load. Body load is the collation of all forces (acceleration/deceleration, change of direction, player-to-player contact and contact with ground) acting on an athlete at a time (9). This can be an important factor (particularly in sports involving player-to-player contact) in creating a more complete picture of the physical demands of a team sport match or session (15). Additionally, integrated accelerometers work independently from the satellite system that drives the positional data recording in GPS (5). This allows impact and body load measures to be recorded both indoors and outdoors, creating substantial value for team sports that use indoor and outdoor training venues (such as professional American gridiron). Kelly et al. (24) tested the validity of accelerometers in a controlled experiment and found the validity of acceleration measures produced by accelerometers was significantly different from criterion measures during static testing (p = 0.001 for both negative and positive axis orientation and 2 devices). There were large percentages of difference to the criterion values, ranging from 27.5% to 30.5% for 4 devices. Similar values were also found during dynamic testing, the accelerometer significantly underestimated peak acceleration by 32–35% when compared against the criterion (24). Although the research is limited it would seem that accelerometers integrated in GPS tracking devices have poor validity for measurements of accelerations during team sport testing. Gastin et al. (14) found when tackles involving 20 professional AFL players wearing 10Hz GPS devices were subjectively analyzed by a group of experts and categorized as either high intensity, medium intensity or low intensity, player load was significantly higher in the high-intensity tackle category than both other categories (p < 0.01), and the medium intensity tackle category was significant higher than the low-intensity tackle category (p < 0.01) (14). Yet, Gastin's et al. (14) use of subjective analysis as opposed to a controlled criterion do little to provide information on the validity of accelerometers measures. However, it does give some evidence that these devices may have acceptable ecological validity for categorizing tackles in Australian football.
Boyd et al. (5) found that accelerometers have a good level of intraunit reliability for measuring player load during static assessment (CV = 1.01%) and dynamic hydraulic assessment at both 0.5g and 3.0g forces (CV = 0.91% and CV = 1.05%, respectively). Similarly, accelerometers have good interunit reliability during static assessment, dynamic hydraulic assessment at both 0.5g and 3.0g forces (CV = 1.10, CV = 1.04, and CV = 1.02%, respectively) and sport-specific assessment (CV = 1.94%) (5). More recently, Kelly et al. (24) found that the intraunit reliability was good (1.87–2.21% CV) when measuring peak gravitational accelerations during impacts created by a purpose built mechanic device. The 4 devices used did not significantly differ (p = 0.484) on measures of peak gravitational acceleration during testing, indicating good interunit reliability (24). In addition, accelerometers have been found to have good to moderate interunit reliability when measuring peak acceleration (g) during 10 m and 30 m sprints, and the frequency of accelerations exceeding 5g during a 10-m sprint (4.69–5.16% CV). However, when measuring frequency of accelerations exceeding 5g during a 30-m sprint, accelerometers seem to have poor interunit reliability (CV = 14.12%) (35).
These findings suggest that accelerometers have very good intraunit reliability and interunit reliability. When using these devices they can be used interchangeably or compared with confidence. Player/body load measurements had noise (CV) less than that of the signal (SWD) meaning when measuring player/body load, accelerometers are able to detect changes or differences in physical activity (5).
All GPS devices are able to accurately quantify distances during team sport simulations and therefore can justifiably be used during team sports matches and training that simulates match play movements. However, all reported 1Hz GPS units and MinimaxX 5Hz GPS units have limitations during short distance linear running. These results need to be acknowledge when tracking player movements during training sessions that are almost entirely comprised these movements (e.g., conditioning drills). Although, there is more recent evidence to suggest that MinimaxX 5Hz GPS units have overcome limitations of short high-intensity running, they still may have problems accurately tracking movement involving frequent changes of direction. Despite this, there have been similar findings in relation to speed and velocity measures. Although all devices seemingly provide accurate measures during sport simulations, during linear running the validity of velocity/speed measure may be compromised. Again this is particularly prominent in 1 and 5Hz GPS units. To accommodate this, it may be preferable when grouping distances based on velocity bands, that broad bands are used to group distance into high velocity and low velocity distance, as opposed to having 5 or 6 velocity bands as some research suggests. From the evidence to date, it would seem that 10Hz GPS units are the optimal GPS tracking device. These devices seemingly overcome limitations of validity seen in 1 and 5Hz GPS units for high-intensity short distance running during. Furthermore, there seems to be no additional benefit to increasing the sample speed to 15Hz, in fact it may be detrimental to measures of distance and speed/velocity. Because of the relatively recent development of the 15Hz GPS devices, there have been relatively few studies, however this may change with time. Likewise, there has been very little research completed to date regarding the validity of load measurements from accelerometers. These measures have very poor validity during controlled testing, although there has been some evidence to suggest ecological validity in Australian football.
Evidence suggests that the intraunit reliability of all devices on almost all occasions is better than the interunit reliability of the device. Therefore, it would be best practice to have athletes wear the same device whenever possible across multiple tracking sessions. This practice would decrease error when comparing results between sessions of a single device. During team sport simulations it seems that the interunit reliability of all devices improves, and therefore, data from separate devices may be compared with relative confidence during a team sport training session or match. However, during both linear/curvilinear and team sport simulated high-speed running reliability worsens, more significantly with interunit reliability. From the literature, it seems that the device that is most reliable is the 10Hz GPS unit. The 10Hz GPS units were able to repeatedly report short distances covered at high velocities with good to moderate intraunit reliability, overcoming limitations of both 1 and 5Hz GPS devices. The 10Hz GPS units also consistently reported distances and velocities with better intraunit reliability than 15Hz GPS devices. Likewise, the 10Hz GPS units had the best interunit reliability for distance and velocities measures across linear, change of direction, and team sport circuits. However, it should be noted that the interunit reliability for high-speed and very high-speed running decreased for the 10Hz GPS units. As mentioned previously, it would still be considered best practice for athletes to wear the same device across multiple sessions or matches. Finally, accelerometers have both very good intraunit reliability and interunit reliability. Again, body load and impact measures recorded by the accelerometer do not rely on interactions with satellites. Therefore, the good reliability of these measures mean comparisons can be made both between and within a device with good confidence (and devices can be used interchangeably) in sports which use both indoor and outdoor training sessions.
Both validity and reliability measures were grouped as good (<5%), moderate (5–10%), and poor (>10%). Although these values are based on previous recommendations for reliability, these values were adapted to validity measures for consistency of reported results and future reporting. As previously mentioned, because of lack of conformity in the literature the comparison of studies required comparing different, although related measures. Readers should take note of what measure they are interpreting and how it affects the interpretation of the statistical score. Furthermore, errors of these magnitudes (<10%) should be seen as acceptable error for most measurements during most team sports. Global positioning systems offer a fast and time effective solution to time motion analysis. It allows for both real-time feedback on player running and immediate reporting once the game or training session has concluded. In contrast, when using video-based time-motion, there is no potential for real-time feedback on athlete movement. Although there have been reports of it taking upward of 8 hours to comprehensively analyze running variables after a session (30). When considering the alternative practices for player tracking, some errors may be an acceptable trade-off for both the time efficient and ease of use of GPS devices. However, it is important to understand that the interpretation of results should come from an understanding of the acceptable level of error in measurement for a specific team sport.
Review of the literature suggests that all GPS units, regardless of sampling rate, are capable of athlete tracking for distance, with adequate intraunit reliability to allow multiple comparisons of a single device. However, coaches and practitioners should be aware of limitations of earlier 1Hz and 5Hz GPS units when interpreting distance during high-intensity running, velocity measures, and short linear running. To date it would seem that 10Hz GPS devices are the most valid and reliable across linear and team sport simulated running, overcoming many limitations of earlier models. Increasingly, the 15Hz GPS devices have had no additional benefit and in fact these performed worse in studies testing both 10 and 15Hz GPS devices.
There were no conflicts of interest among any authors. There was no funding for this research. The results of this study do not constitute endorsement of any product by the authors or the National Strength and Conditioning Association.