Comparison of GPS and IMU systems for total distance, velocity, acceleration and deceleration measurements during small-sided games in soccer

Abstract Physical variables from soccer training are commonly quantified by using global positioning systems (GPS) and/or inertial measurement unit (IMU) systems. More studies are needed to compare both systems, taking into account that both are frequently used to measure physical load. The purpose of this study was to (i) compare IMUs (PlayerMaker™) to ten Hz GPS (Polar Team Pro) units during a square running protocol (SRP) to investigate the accuracy of distance measured and intra-unit reliability of the devices and (ii) compare IMUs and GPS units for measurements of total distance (TD), very high intensity running (VHIR), accelerations and decelerations during small-sided games (SSGs) in semi-professional soccer. Four subjects ran theSRP whilst wearing two IMU units and three GPS units to compare the accuracy of distance measured and examining the intra-unit reliability. Fifteen male soccer players participated in the SSGs where the players’ physical variables were simultaneously captured by a foot-mounted IMU in addition to GPS units. IMU and GPS were respectively 2% and 1% from actual distance, whereof both systems had a coefficient of variation at~1%. In the SSGs, the IMU measured~4% higher TD compared to the GPS. VHIR measured by the IMUs were around 18% lower while the IMUs measured~39% more accelerations and~35% more decelerations compared to the GPS. Both systems measure close to actual distance. In SSGs, IMU measures higher distances below VHIR while the GPS measures higher distance above VHIR. IMUs measure a higher number of accelerations and decelerations during SSGs.


Introduction
In recent years, there has been a substantial increase in research aimed at measuring and analyzing physical performance demands and training load in soccer (Cummins et al., 2013;Dalen et al., 2016Dalen et al., , 2019Iaia et al., 2009;Sarmento et al., 2014;Warr et al., 2020). A variety of sport tracking devices from a range of manufacturers can provide valid and reliable data instantaneously, including locomotion categories (e.g., high intensity running and sprinting), accelerations and decelerations from training and matches (Cummins et al., 2013;Girginov et al., 2020;Whitehead et al., 2018). Time motion analysis data have reported that soccer players often will cover a total distance (TD) of 10-14 km during a match (Bradley et al., 2013;Dalen et al., 2016), whereas 2500-2800 meters is in running (>14.4 km · h −1 ) (Iaia et al., 2009) and approximately 800 meters in high-intensity running (HIR) (>19.8 km · h −1 ) (Dalen et al., 2016;Iaia et al., 2009). Around 300 meters is sprinting (>25.2 km · h −1 ) (Bradley et al., 2013) and the players perform on average 76 accelerations during a 90-minute match (Dalen et al., 2016(Dalen et al., , 2019. Typically, the external training load is quantified by using microtechnology systems such as global positioning systems (GPS) and/or micro-electrical mechanical systems (MEMs, e.g., accelerometer) (Whitehead et al., 2018;Züll et al., 2019). Global positioning system (GPS) is a satellite-based navigation system, and the use of GPS in team sports is therefore limited to outdoor use (Huggins et al., 2020). The manufacturers typically deliver GPS units with a measuring frequency of either 1, 5, 10, 15 or 18 Hertz (Hz) (Huggins et al., 2020), of which all have been reported with good validity (Aughey, 2011;Beato et al., 2018;Coutts & Duffield, 2010;Hoppe et al., 2018;Jennings et al., 2010;Johnston et al., 2014;Rampinini et al., 2015) and reliability (Aughey, 2011;Beato et al., 2018;Castellano et al., 2011;Coutts & Duffield, 2010;Hoppe et al., 2018;Huggins et al., 2020;Jennings et al., 2010;Scott et al., 2016;Varley et al., 2012) when it comes to total distance covered in an activity. The accuracy and precision of velocities, however, seems to be limited and is to a greater extent dependent on the sampling frequency, of which a higher sampling rate gives a greater validity (Hoppe et al., 2018) and reliability (Hoppe et al., 2018;Varley et al., 2012). Especially, ten Hz GPS units have been reported to have very good validity when it comes to distance covered (Scott et al., 2016), consisting of distance validity on short sprints (Castellano et al., 2011), specific movements in field-based team sports activities (Johnston et al., 2014;Vickery et al., 2014) and also high intensity running in field-based team sports (Rampinini et al., 2015). Nevertheless, studies have reported that GPS units provide a satisfactory validity of top speed during field sport activities (Beato et al., 2018;Vickery et al., 2014). Consensus among authors is still that the validity of GPS units during field sports activities decreases with increasing speed of the athletes (Aughey, 2011;Rampinini et al., 2015). Furthermore, use of GPS units is suggested to have limitations for assessment of short and high speed efforts with changes of direction (Jennings et al., 2010;Vickery et al., 2014). Johnston et al. (Johnston et al., 2014) suggest that ten Hz GPS units have the ability to measure distance during linear running and also team-sport field activities with varying speeds and different distances with a high degree of validity. The same author suggests satisfactory validity of ten Hz GPS units for measuring instantaneous velocities during constant velocity running and also running involving accelerations, of which the validity increases proportionally with a higher initial velocity in the activity. Vickery et al. have reported good validity of velocity during short sprints including accelerations and also during a field-based sport movement protocol (Akenhead et al., 2014;Vickery et al., 2014). These findings are supported by Akenhead et al. (Akenhead et al., 2014) who found that instantaneous velocity measures were valid when the acceleration occurred between 0-4 m·s −2 . However, instantaneous velocity measures were not valid when acceleration was above 4 m·s −2 (Akenhead et al., 2014). Some GPS devices have previously been delivered with an integrated inertial measurement unit (IMU). The IMUs consist of accelerometers, gyroscopes, and can also include magnometers with the intention to provide a complementary method for monitoring the external training load such as "Player effort" and "Player Load TM " (Barreira et al., 2017;Barrett, 2017;Boyd et al., 2011). Recently, IMUs have increasingly also been used as an alternative tool for measuring locomotion categories via the detection of the individual's stride length, frequency and running pattern (Godfrey et al., 2015;Potter et al., 2019), and can be fitted to monitor physical performance in team sports by tracking velocity-based variables (Van Der Kruk & Reijne, 2018). By measuring accelerations and angular velocity of the athlete, changes in gait events may be detected for each stride (Yang et al., 2011) and determine velocity and orientation of various body parts (Hausswirth et al., 2009;Waldron et al., 2021). A comparison of foot-mounted IMUs to two 10 Hz GPS devices from separate suppliers found that IMUs measured a significantly higher TD during a soccer specific running protocol compared to the GPS, which was attributed to greater distance in velocities below HIR. One out of the two ten Hz GPS suppliers measured a significantly higher distance compared to the IMUs in velocity zones above HIR, while no significant differences were found in distance measured above HIR from the other GPS supplier and the IMUs (Dalen et al., 2019;Waldron et al., 2021). Furthermore, Waldron et al. (Waldron et al., 2021) reported no differences when measuring peak velocity between IMUs and ten Hz GPS. When comparing the mean velocity on the other hand, the IMUs measured significantly higher than the ten Hz GPS (Waldron et al., 2021). Van der Kruk and Reijne (Van Der Kruk & Reijne, 2018) suggest that IMUs potentially could replace the more traditionally used GPS units for measuring velocity-based metrics in soccer, knowing that the IMU does not depend on satellite signals. Further reasoning argues greater applicability due to IMU availability to be used indoors as much as outdoors (Van Der Kruk & Reijne, 2018).
Quantification of time motion-variables during soccer is essential for training load monitoring and control. Elite sport environments are continually looking for improvements and are always aiming to optimize the training process. The ability of player-tracking devices to accurately and reliably quantify movement may be an important factor for training outcomes, and high quality measurements of physical variables in soccer is beneficial for coaches in the planning and implementation of appropriate training sessions (Živanović et al., 2011). Otherwise, coaches may need to manage the training load from insufficient measurements, which could result in inappropriate training load. A training load too low will lead to a lack of progress while a training load too high often will cause injuries among the players (Fornasier-Santos et al., 2021;Impellizzeri et al., 2020). In light of this, high intensity bouts such as accelerations, decelerations and running in high velocities are reported to be crucial in the determination of the external training load in soccer and should therefore be taken into consideration (Barreira et al., 2017;Iaia et al., 2009;Robey et al., 2014). SSG in training is a well-known training method with the aim of creating an overload on physical demands (in addition to the technical and tactical aspects) in soccer (Dellal et al., 2012;Halouani et al., 2014;Moreira et al., 2016). The gathering of valid and reliable metrics from relevant and soccer-specific training is therefore of high importance, and there is a need for further research on this topic. Utilization of technology is more and more relevant in the society (Madl et al., 2021) in general, and naturally as well also in the soccer community.
There is a lack of studies comparing foot-mounted IMUs and ten Hz GPS devices to each other. Also, little research addresses the possible differences in data output from soccer specific movements with the purpose of discussing the impact it has on determining training load. Initially, this study will compare IMU and GPS measurements of distance and velocity with each other and up against a criterion measure. Continuing, we will investigate differences in data output from the two devices on soccer specific movements during SSGs. The purpose of this study is therefore to compare two different methods to measure physical variables in soccer: a foot-mounted inertial measurement unit to a ten Hz global positioning system unit for the measurement of total distance covered, velocities, accelerations and decelerations during small-sided games in semiprofessional soccer. Based on previous research we hypothesized that both systems would measure close to total distance against a criterion measure during a running protocol. During the small-sided games protocol we hypothesized that the two devices would measure total distance with relatively little variety, while we were expecting a greater variety between the two devices during high intensity variables such as high-intensity running, sprinting, accelerations and decelerations. To our knowledge, we are the first study to compare these two measuring units during small-sided games.

Subjects and samples
Initially, in this study, four subjects volunteered to perform a square running protocol (N = 4; mean ± SD: age = 34.5 ± 9.57 years [range 26-48 years old], height: 182 ± 2.82 cm, weight: 77.25 ± 4.34 kg). The square running protocol (SRP) consisted of five laps of 241 meters per lap, where the participants were running together with two meters distance between each other on the white lines on one half of a normal sized artificial soccer pitch (goal line/halfway line: 68 meters, sideline: 52.5 meters, duration: 7 minutes and 36 seconds). The SRP define the criterion measure in this study of which the known distance is 1205 meters. The data from the SSGs were collected from semi-professional male soccer players, competing in the Norwegian second level (N = 15; mean ± SD: age = 25 ± 3.36 years [range 20-32 years], height: 182.1 ± 5.1 cm, weight: 74.9 ± 3.4 kg). Total distance, distances in different velocity zones, peak velocity, sprints, accelerations and decelerations were collected and analyzed from playing nine games of 5 vs. 5 + goalkeeper SSGs with a duration of approximately 2.5 minutes per game. Some measurements were excluded due to measurement errors but apart from those, individual measurements of each player from every SSG were collected (N = 69). The distance measurements from each game were divided by minutes of playing time, which means that results are reported per minute of playing time. Goalkeepers were excluded from the study. All games were played outdoors on artificial grass during stable weather conditions (6° Celsius, 0,0 mm rain/snow, 3 m/s wind and 47% humidity). The games were conducted in the middle of the pitch which may be beneficial for minimizing possible exterior disturbances interfering with the GPS signal. All games were played on a 40 × 32 meter pitch, giving 107 m 2 per player (including goalkeeper).

Data collection
During the SRP, each of the participants were fitted with two sets of foot-mounted inertial measurement units (IMU, PlayerMaker™), and three GPS (Polar Team Pro) units with the purpose to increase the number of measurements in addition to examining the intra-unit reliability. The PlayerMaker™ system measures at 1000 Hz and the sensor consists of a triaxial accelerometer and a tri-axial gyroscope (MPU-9150, InvenSense, California, USA) for measurements of velocity, and also measurements of rapid changes in velocities of each foot during gait. The PlayerMaker™ system can measure soccer-specific variables through a custom-built machine learning algorithm and is calculated from players' gait cycles during activity. The resulting output can provide relevant metrics, such as distance covered during soccer-specific exercise. In addition to the double set of IMU, the players also used three global positioning system (GPS) units (Polar Team Pro) fastened in body-worn belts around the participants' chests. The participants were wearing the belts directly on their skin and the sensors were placed in the middle portion of the front of the chest, just below the sternum. The GPS system uses satellite signals to locate the players' locations and movements on the pitch. The positional data was sampled at 10 Hz. For measuring distances, velocities, accelerations and decelerations during SSGs, the players wore a foot-mounted IMU sensor which is part of the PlayerMaker™ system. Each player had one unit attached to each of their feet which was held by a tight custom silicon strap where the sensor was located on the lateral aspect of the calcanei (Waldron et al., 2021). In addition to the set of IMU, the players also used one global positioning system (GPS) unit (Polar Team Pro) fastened in body-worn belts around the their chests during the SSG protocol. Previous studies have reported good validity and reliability of ten Hz GPS units for soccer specific exercise (Castellano et al., 2011;Scott et al., 2016).

Procedures
The players' total distance covered, distances in velocity zone (VZ) 1(<7.19 km · h −1 ), 2(7.2-14.39 km · h −1 ), 3(14.4-19.79 km · h −1 ), very high intensity running ((VHIR) >19.8 km · h −1 ), peak velocity (PV) and sprints (>25.2 km · h −1 ) were measured during small-sided games (SSG, 5v5 + goalkeeper) in training. Furthermore, numbers of accelerations in acceleration zones (AcZ) 1(<2 m · s −2 ) and 2 (≥2 m · s −2 ) were measured along with numbers of deceleration in deceleration zones (DecZ) 1(>-2 m · s −2 ) and 2(≤-2 m · s −2 ). AcZ2 and DecZ2 correspond with acceleration and deceleration thresholds reported in previous studies (Dalen et al., 2016(Dalen et al., , 2019. The distance covered in each velocity zone are expressed as meter per minute of play, and accelerations and decelerations are expressed as numbers per minute of play, while the peak velocity, and measurements during the SRP are presented as means of the absolute values for each measurement unit the participants wore. The players' distances, velocities, accelerations and decelerations were simultaneously captured by a foot-mounted IMU (PlayerMaker™) fastened around the players' soccer shoes in addition to GPS units (Polar Team Pro™, version 1.4.2) worn on belts around the players' chests. TD and distances in VZ1, VZ2, VZ3, and VHIR were measured during both the SRP and the SSGs. Velocity thresholds in this study are similar to previous studies, which means that VHIR is analyzed as the sum of distance above the minimum velocity of HIR (>19,8 km · h −1 ) (Dalen et al., 2016(Dalen et al., , 2019Iaia et al., 2009). Additionally, the PV measured and number of sprints were included in this study. To be counted as a sprint the players had to reach a speed of>25.2 km · h −1 for≥1 second.

Statistical analysis
All statistical analyses were performed in IMB SPSS Statistics version 27.0 (New York, USA). Descriptive statistics were conducted to find means and standard deviation for the different categories. Normality of the differences between the two measurement units was assessed for each outcome variable with Shapiro-Wilk tests (all p ≥ 0.05) as well as visually (histograms, Q-Q plots). Systematic comparison of TD, VZ1-3, VHIR, PV, AcZ1, AcZ2, DecZ1 and DecZ2 between the two measurement units was investigated by performing paired sample t-tests for the different categories. For total distance and average velocity, comparison to the criterion measure was done with one-sample t-tests. Bland-Altman analysis were conducted to highlight the difference between the IMUs and GPSs measurements of total distance during the SRP. Because of the higher number of GPS units during the SRP, two of the three units were randomly selected for each participants and included in the Bland-Altman plots for comparison with the two IMU. Coefficient of variation (CV) was used to investigate the intra-unit reliability for each system in the SRP, while intraclass correlation coefficient (ICC) estimates and their 95% confidence intervals (CI) were calculated based on a two-way mixed model and an absolute-agreement type. Statistical significance was accepted at p < 0.05.

Results
During the SRP both measurement systems showed a slightly lower absolute TD compared to the actual measured distance. The difference between the mean TD for IMU and actual distance was about 2%, which corresponds to 25 meters of 1205 meters in total. The corresponding measurement for GPS was about 1% which equates to 14 meters of 1205 meters in total (see Table 1). Moreover, both systems had a coefficient of variation (CV) at~1% in the measurements of TD (see Figure 1). Comparison between IMU and GPS units during the SRP showed lower measurements in VZ2 (~2.5%), PV (~2.7%), AcZ2 (~39%), DecZ1 (~59%) and DecZ2 (~83%) for the IMUs. No differences were found between IMU and GPS units for TD during the SRP (Table 1). In the SSGs, the IMU measured about 4% higher total distance compared to the GPS unit (see Table 2 and Figure 2A). Distance in VZ2 and VZ3 was~6% and~8% higher for the IMUs compared to the GPS units, respectively (see Table 2, Figure 3A). Furthermore, distances in VHIR measured by the IMUs were around 18% lower than the GPS units (see Table 2, Figure 3A). The IMUs measured a higher AcZ2 (~39%) and a higher DecZ2 (~35%) compared to the GPS (see Table 2, Figure 3B). DecZ1, however, showed a lower (25%) measurement when compared to the GPS (see Table 2 and Figure 3B). Finally, no significant differences were found in VZ1, PV, AcZ1 or DecZ1 between the two measuring devices (see Table 2, Figures 2B and 3A).

Discussion
The first finding in this study was that in the SRP measurement from both the IMU and GPS it was close to the actual distance, whereas the GPS measured a slightly higher distance (~10 meters) than the IMU. The participants were running at a steady pace with an average speed of 9.51 km · h −1 during the SRP which suggests that both measuring devices measure close to actual distance in VZ2 during this protocol. We assume this to be likely, despite taking into account changes of direction during the SRP (with deceleration and acceleration). The high degree of accuracy of the two devices during this protocol may help coaches to have confidence in the training data output when measuring TD, and provide a more convenient use of the measuring devices in the training process. Also, the SRP in this study presents a satisfactory intra-unit reliability with a relatively low CV (~1%) for both the IMUs and GPS.
However, in the SSGs this study found that IMUs measured a significantly higher TD per minute of play during SSG compared with the GPS units. The~4% higher TD for the IMUs from SSGs is because of a significantly higher distance in VZ2 and VZ3. In addition, the IMU also had higher values of accelerations and decelerations above 2 m · s −2 than the GPS units during SSGs. Findings of higher TD measured by IMUs are supported by Waldron et al. (2021), who reported a higher TD compared to the GPS during a soccer specific running protocol, which was due to a considerably higher distance in velocity zones below VHIR (>19.8 km · h −1 ) (Waldron et al., 2021). This present study on the other hand compares measurements of the two devices during actual soccer play which may contribute to a greater practical understanding for the data output in training. Furthermore, the IMUs measured significantly lower distances in VHIR compared to the GPS, even though the total meters in these zones during the SSGs are scarce. SSGs in soccer leads to a great amount of frequently high-intensity actions and requires a high degree of precision and accuracy to collect valid and reliable metrics from the players. In the light of this, previous studies have reported good validity on distance measurements from the use of GPS units for various velocities during soccer specific activities similar to this present study (Castellano et al., 2011;Rampinini et al., 2015;Scott et al., 2016). At the same time, authors have also reported gradually lower validity of GPS units with increasing speed of the players (Aughey, 2011;Rampinini et al., 2015) which may question this current finding. On the other hand, the results in this study showed that a very low proportion of the TD actually came from distance covered in VHIR for both GPS and IMU. This may be due to the SSG format, with a relatively small pitch size (107 m 2 per player), which is not especially suited for facilitating high velocity runs such as HIR and sprinting (Dalen et al., 2019).
Another major finding in the present study was that the IMUs measured a significantly higher number of accelerations and decelerations above 2 m · s −2 per minute of play compared to the GPS. This finding shares similarities with Waldron and colleagues (2021) who reported a significantly higher number of accelerations and decelerations higher than 1.5 m · s 2 for the IMU compared to the GPS (Waldron et al., 2021). Despite the similarities, the different acceleration and deceleration thresholds in this study make the comparison only an assumption. GPS units have been reported with weaknesses in assessment of short and high-speed Note: † = Significantly lower than actual distance (p < 0.001), ‡ = Significantly lower than actual distance (p < 0.05), * = IMU significantly lower than GPS (p < 0.05). Data are presented as mean ± standard deviation. n = numbers, m = meters. CV = coefficient of variation. ICC = Intraclass Correlation Coefficient. LL = Lower limit. UP = Upper limit. CI = Confidence Interval. TD = Total distance covered (m); AV=Average velocity (km · h −1 ); PV = Peak velocity (km · h −1 ). SRP = Square running protocol, IMU = Inertial measurement unit, GPS = Global positioning system efforts with change of directions (Jennings et al., 2010), which may argue that GPS units are not suitable for detection of accelerations and decelerations during SSGs in soccer. The footmounted IMUs have a higher sampling rate (1000 Hz) than the GPS units and there is also an major difference in the anatomical placement of the two measuring units. Nevertheless, there is no research that provides any evidence of one being better that the other in terms of measuring accelerations and decelerations during SSGs. Furthermore, the lack of a criterion measures during this protocol (5 vs. 5 SSGs) makes it inappropriate to favor one over the other to assess these variables. Anyway, the lack of studies investigating the validity of footmounted IMUs during SSG in soccer leads to uncertainty on the topic, after ten Hz GPS units have been reported to have satisfactory validity of accelerations during soccer specific activities (Akenhead et al., 2014;Halouani et al., 2014). Akenhead et al. (2014) reported that instantaneous velocity measures of ten Hz GPS units were valid when the acceleration occurred between 0-4 m·s −2 (Akenhead et al., 2014), which extend beyond the acceleration threshold in this present study.

Perspectives and practical applications
The results in this present study show significant differences between the two measurement systems during 5 vs. 5 SSGs for VHIR, accelerations and decelerations. These physical parameters may be considered as crucial variables in the determination of the external training load in soccer, and are especially important to avoid possible injuries (Barreira et al., 2017;Dellal et al., 2012;Impellizzeri et al., 2020). In the process of managing the training load among the players, coaches therefore need to be aware of to what degree the training data output from the two different measurement systems gives acceptable accuracy. It is therefore essential to investigate the training data output from actual soccer play in training, and based on the results in this study the practical consequences could hypothetically be substantially different based on which one of the two tracking devices is being used. Hence, a coach's knowledge regarding differences in validity and reliability of GPS units with different Hz frequency as well as IMUs is apparently very important. Therefore, a critical view on training data output from both IMU and GPS is suggested. Due to the results of the present study coaches are advised to be selective when using previous studies as a reference on players' performance of VHIR, accelerations and decelerations during SSGs in soccer. Furthermore, we will recommend coaches learn about the velocity and acceleration/deceleration thresholds for the different categories (e.g., HIR, sprinting, etc.) for the different measurement systems. Quantification of training data above or below essential thresholds may give major practical consequences for the training process, despite relatively low differences in the measurement values. Additionally, we would like to suggest vigilance in determining physical demands in SSGs in soccer by comparison of training load output from other sources. This is due to a lack of studies investigating the inter-and intra-unit reliability of these different measuring devices during SSG in soccer. In the process of managing the training load, players should therefore only be compared to themselves and use the same measuring unit from one time to another, preferably continuously over time to collect specific and relevant information from each player. Note: # = IMU Significantly higher than GPS (p < 0.001), § = IMU significantly higher than GPS (p < 0.05), * = IMU significantly lower than GPS (p < 0.05). Data are presented as mean ± standard deviation per minute of play. TD = Total distance covered; VZ1 = Velocity Zone 1 (<7.19 km · h −1 ); VZ2 = Velocity Zone 2 (7.2-14.39 km · h −1 ); VZ3 = Velocity Zone 3 (14.4-19.79 km · h −1 ); VHIR = Very high intensity running (>19.8 km · h −1 ); PV = Peak velocity; Sprints (>25.2 km · h −1 ). AcZ1 = Acceleration Zone 1 (<2 m · s −2 ); AcZ2 = Acceleration Zone 2 (>2 m · s −2 ). DecZ1 = Deceleration Zone 1 (>-2 m · s −2 ); DecZ2 = Deceleration Zone 2 (<-2 m · s −2 ). m · min −1 = meters per minute, km · h −1 = kilometers per hour, n · min −1 = numbers per minute. IMU = Inertial measurement unit, GPS = Global positioning system Ten Hz GPS have documented acceptable validity for measuring physical variables during soccer specific exercise, which strengthens these systems' credibility. IMUs may be a good alternative to the GPS but need more research.

Conclusion
This study is the first to compare a ten Hz GPS with a foot-mounted IMU for measurements of physical variables in SSGs in soccer. Firstly, both IMU and GPS devices measure close to actual distance during the SRP. In SSGs, the results showed that TD were significantly higher for the IMU and that the IMU measures significantly higher distances in lower velocities than the GPS. The GPS measures significantly higher distances in high velocities. Furthermore, IMUs measure a significantly higher number of accelerations and decelerations during SSGs.

Figure 3. Players' mean and SD values from GPS and IMU of distance covered in different velocity zones (A) and accelerations and decelerations (B)
, per minute of play in 5vs5 SSGs. # = IMU Significantly higher than GPS (p < 0.001), § = IMU significantly higher than GPS (p < 0.05), * = IMU significantly lower than GPS (p < 0.05). Data are presented as mean ± standard deviation per minute of play. m · min −1 = meters per Figure 2. Mean total distance covered per meter (A) and peak velocity (B) during 5vs5 SSGs measured by GPS and IMU units. # = IMU Significantly higher than GPS (p < 0.001). Data are presented as mean ± standard deviation per minute of play. m · min −1 = meters per minute, km · h −1 = kilometers per hour, IMU = Inertial measurement unit, GPS = Global positioning system, SSGs = Small-sided games.

Data availability statement
The datasets generated and analyzed for this study can be requested by correspondence author at simen.r.sand-mal@nord.no.

Citation information
Cite this article as: Comparison of GPS and IMU systems for total distance, velocity, acceleration and deceleration measurements during small-sided games in soccer, Simen Sandmael & Terje Dalen, Cogent Social Sciences (2023), 9: 2209365.