Article

Placement of team sport GPS devices for reliability assessment

Authors:
  • Unemployed
To read the full-text of this research, you can request a copy directly from the authors.

Abstract

The goal of this study was to determine the effect of unit placement on signal integrity and measurement output obtained during reliability trials for global positioning system (GPS) devices. Two of the same model GPS units were worn in four separate unit configurations (Piggyback, Above-Below, Front-Back and Lateral) during 4 × 2 lap repeats of a team sport simulation circuit. Differences in signal integrity indicators (# satellites, horizontal dilution of precision (HDoP), signal-to-noise ratio (SNR) and signal strength (SS)), total distance covered in each lap, and distance covered above and below 3 m·s ⁻¹ , were compared between the two units. The results showed that for signal integrity measures, differences between units were negligible for Above-Below and Lateral, but comparatively high for Piggyback and Front-Back. For distance measures, values were similar between units for Above-Below, however, discrepancies occurred in both total distance and speed distribution for the other configurations. This study concluded that the Above-Below unit configuration yielded the smallest differences between units for signal integrity and measurement output. Therefore, the Above-Below configuration is recommended for future GPS reliability investigations.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

... Each participant wore two sensors during the session. The first sensor was attached to the posterior side of the participants' shoulder pads, at the midline (spine) of the body and near the superior 1/3 of the scapula, while the second sensor was directly above it (as suggested in GPS reliability research by Polglaze, Tan, & Peeling, 2021). Communication between anchors and the player sensors (through ultra-wideband channels ranging from 3244.88-4742.40 ...
... Specific to LPS, reliability of sensors is reduced with short and sudden changes in velocity (Crang et al., 2021). Regarding sensor location, the first sensor was attached to the posterior side of the participants' shoulder pads near the superior 1/3 of the scapula and the second sensor was attached directly above it (within 5 cm), which have been reported as the preferred sensor locations for testing the reliability of GPS (Polglaze et al., 2021). Manufacturer guidelines also outline the importance of sensors being secured in the same location for each use to ensure reliable and valid results. ...
Article
Full-text available
This study determined the reliability and validity of a Kinexon local positioning system (LPS) for measuring external load in ice hockey players during an on-ice session. Fourteen ice hockey players (25.1 y, 78.6 kg, 176.9 cm) wore two LPS sensors to examine the inter-sensor reliability of the LPS during an on-ice session, and LPS speed and acceleration were measured during 40 m linear on-ice sprints and compared to a previously validated robotic sprint device to examine LPS accuracy. The coefficient of variation (CV), standard error of measurement (SEM), and intra-class correlation coefficient (ICC) were calculated for each LPS measure. Pearson's correlations, simple linear regressions, and Bland-Altman plots were used to test the agreement and relationship between the two systems. Statistical significance was determined at p < 0.05. The majority of LPS measures were reliable (CV < 10% and ICC > 0.9) when comparing the two sensors worn by each player. Peak speed, speed at 5 m, and 0-5 m acceleration were all comparable to those reported by the robotic sprint device, with nearly perfect (peak speed and 0-5 m acceleration) and very large (speed at 5 m) magnitudes of correlation and mean biases <0.5 km/hr for speed measures and <0.01 m/s2 for acceleration. The present results demonstrate that the Kinexon LPS is reliable and accurate for investigating on-ice external load in ice hockey players when sensors are consistently secured on the back of the players' shoulder pads.
... The parameters characterizing the external loads were: total distance (TD) [ In addition to the sampling frequency of the device (10Hz), the accuracy of positioning measurement and its derivatives is influenced by other factors. The most important of these include the number of satellites connected and their horizontal dilution of precision (HDOP) [22,23]. HDOP values >1 are defined as ideal, while values >2 are defined as excellent [24]. ...
Article
Full-text available
Amputee Football (AF) players move using lofstrand crutches (LC) and a single leg during training and matches, which may expose them to excessive loads. Due to a lack of scientific articles describing these issues, this pilot study aims to gain insight into the assessment of pain experiences, as well as external loads during training and matches in AF. An observational study design was followed. Twelve male AF players (2 goalkeepers—GK and 10 field players—FP), 29.9 ± 8.7 years, stature 178.3 ± 6.5 cm, body mass 77.2 ± 8.9 kg were involved in the study. After stature and body mass measures, participants filled out the questionnaire for pain. The players were familiarized with digital pain drawing (DPD). Afterwards, they reported pain typically felt in relation to regular AF training and playing matches—recall pain (RP). During a two-day training camp (TC) with two training sessions each day and during a two-day international tournament (IT) with two matches each day, the players were monitored using DPD, rate of perceived exertion (RPE) and overall pain level (OPL) scales. In addition, during an international tournament (IT), match players were monitored using a GNSS tracking system for external load assessment. All of the participants reported multiple locations of pain after AF training or a match. The area of recall pain (RP) was the highest: +0.5% when compared to the end of TC and +43% when compared to end of IT. The pain area registered at the end of IT was significantly lower (p = 0.028) compared to RP and lower without statistical significance when compared to the end of TC. Average RPE was 3.31 ± 1.38 and average OPL was 2.86 ± 1.81 in 0–10 scale. Typical RPE was higher than that registered at the end of IT. Also typical OPL was higher compared to that registered after the end of TC, which was higher than after the end of IT. The average distance covered by a FP during a match ranged from 2483.14 ± 583.64 m to 2911.08 ± 828.90 m. AF field players suffered pain as a consequence of training and matches. The loads coming from playing and training, combined with pain, may lead to injuries. Further research directions should include assessments of the relationship of pain characteristics, injuries and GNSS tracking parameters.
Article
The purpose of this study was to determine the validity and reliability of the GPExe pro2 device to identify movement orientation in team-sport locomotion. Part A involved a series of discrete movement tasks of varying direction and orientation at two speeds (walk, run); Part B involved a continuous Simulated Game Circuit (SGC). For validity, distance in each orientation quadrant (forward, backward, left, right) and two aggregated sectors (lateral, non-forward), was compared to the known distance and calculated as the standard error of estimate (SEE). Inter-unit reliability was assessed in the SGC and reported as typical error (TE) and coefficient of variation (CV). In total, 256 trials were collected for validity (Part A 232, Part B 24), and 17 trials for reliability. Validity improved with faster and more continuous movement (SEE: walk 6.2–8.7%; run 4.4–8.2%; SGC 1.0–6.2%), was better for forward and backward (1.0–2.8%) compared to left and right orientation (3.8–6.2%), and for aggregated sectors (2.4–5.1%) rather than specific quadrants. Reliability was good (CV<5%) for forward and backward quadrants and both aggregated sectors, but poor (CV>10%) for left and right quadrants. These devices provide valid and reliable movement orientation data for team-sport locomotion, particularly in differentiating between forward and non-forward movement.
Article
Full-text available
This study compared the outputs of three different commercially-available GPS player-tracking devices for a range of commonly used displacement and energetic variables in activities replicating team sport movements. Professional male soccer players (n = 7), simultaneously wore three GPS devices (Catapult OptimEye S5, GPExe Pro 1, StatSport ViperPod) whilst completing four separate drills, comprising progressively more complex changes in speed and direction. Displacement (distance, speed) and energetic (energy cost, metabolic power, energy expenditure) variables were compared for each device. All three devices tended to under-estimate distance compared to the known value for each drill, with only minor and inconsistent differences between devices. There were no differences between devices for average speed. For energetic variables, substantial differences were found between each device, and these differences magnified as movement tasks became more erratic. Given that energetic variables are derived from measures of instantaneous speed, and also incorporate the magnitude and direction of change between successive data points, these differences may be attributable to disparities in raw data quality, filtering techniques and calculation methods. In order to provide comparable estimates of energetic variables in team sports, player-tracking devices must be capable of accurately recording instantaneous velocity in activities comprising frequent changes in speed and direction.
Article
Full-text available
Purpose: To compare data from conventional GPS and new GNSS-enabled tracking devices, and to examine the inter-unit reliability of GNSS devices. Methods: Inter-device differences between 10 Hz GPS and GNSS devices were examined during laps (n=40) of a simulated game circuit (SGC) and during elite hockey matches (n=21); GNSS inter-unit reliability was also examined during the SGC laps. Differences in distance values and measures in three velocity categories (low <3 m.s-1; moderate 3-5 m.s-1; high >5 m.s-1) and acceleration/deceleration counts (>1.46 m.s-2and < -1.46 m.s-2) were examined using one-way ANOVA. Inter-unit GNSS reliability was examined using the coefficient of variation (CV) and intra-class correlation coefficient (ICC). Results: Inter-device differences (P <0.05) were found for measures of peak deceleration, low-speed distance, % total distance at low speed, and deceleration count during the SGC, and for all measures except total distance and low-speed distance during hockey matches. Inter-unit (GNSS) differences (P <0.05) were not found. The CV was below 5% for total distance, average and peak speeds and distance and % total distance of low-speed running. The GNSS devices had a lower HDoP score than GPS devices in all conditions. Conclusions: These findings suggest that GNSS devices may be more sensitive than GPS in quantifying the physical demands of team sport movements, but further study into the accuracy of GNSS devices is required.
Article
Full-text available
Athlete tracking devices that include global positioning system (GPS) and micro electrical mechanical system (MEMS) components are now commonplace in sport research and practice. These devices provide large amounts of data that are used to inform decision-making on athlete training and performance. However, the data obtained from these devices are often provided without clear explanation of how these metrics are obtained. At present, there is no clear consensus regarding how these data should be handled and reported in a sport context. Therefore, the aim of this review was to examine the factors that affect the data produced by these athlete tracking devices to provide guidelines for collecting, processing, and reporting of data. Many factors including device sampling rate, positioning and fitting of devices, satellite signal and data filtering methods can affect the measures obtained from GPS and MEMS devices. Therefore researchers are encouraged to report device brand/model, sampling frequency, number of satellites, horizontal dilution of precision (HDOP) and software/firmware versions in any published research. Additionally, details of data inclusion/exclusion criteria for data obtained from these devices are also recommended. Considerations for the application of speed zones to evaluate the magnitude and distribution of different locomotor activities recorded by GPS are also presented, alongside recommendations for both industry practice and future research directions. Through a standard approach to data collection and procedure reporting, researchers and practitioners will be able to make more confident comparisons from their data, which will improve the understanding and impact these devices can have on athlete performance.
Article
Full-text available
The knowledge about physical demands in different sports has increased, thanks to the application of global positioning system devices. The reliability and validity of 10 Hz global positioning system devices have been assessed by some authors. The majority of the studies only addressed the reliability of the devices or, in other words, the ability of scores of global positioning system device to differentiate among subjects or objects. The reliability is based on correlations (such as the intraclass correlation coefficient) which do not give the researcher information that can be interpreted in a practical way. In this way, the aim of this study was to assess the grade of agreement among repeated measurements made on the same subject using two global positioning system devices simultaneously. Four trained male tennis players participated in the study. The participants completed tennis-simulated point-games (n = 32), each player wearing two devices at the same time. Global indicators, such as Player Load (PL), Exertion Index (EI) and Equivalent Distance Index (EDI) per minute, were monitored through the use of global positioning system devices (MinimaxX v4.0; Catapult Innovations, Melbourne, Australia) operating at the above-mentioned sampling frequency of 10 Hz. The systematic error is that there is tendency of the global positioning system devices to measure systematically different from others Vmean (−1.03 m · min−1), Vpeak (−10.31 m · min−1), Equivalent Distance Index (0.63 ratio), PLmin (0.35 UA min−1) and EImin (−0.01) variables. As for random error (limit of agreement), we would expect that in PLmin, the global positioning systems would differ in 95% of the cases between 2.12 and −1.42 m · min−1; any value out of the limits of agreement would result relevant for the practical point of view. We concluded that the global positioning system devices produce systematically different results from one another; therefore, the bias from one global positioning system to another should be subtracted to compare the results between the global positioning systems.
Article
Full-text available
Over recent decades, the use of player tracking technology to monitor physical work output has become established practice in many team sports. Early tracking systems were manual in nature, relying on subjective assessments and arbitrary classifications of movement intensity. Poor spatial and temporal resolution meant that only gross displacement measures could be used to infer energy demands. However, the advent and evolution of automated systems, with higher sampling rates and improved accuracy, have enabled data collection to occur on a mass scale, and served as a catalyst for extensive research into the demands of team sport activity, including comparisons between different groups of athletes, and the effects of various interventions on performance. The inherent assumption with this research is that, based on steady-state models where energy cost is independent of speed, total distance and average speed are indicative of the amount and rate of work done, respectively. This assumption could be justified if the activity was performed at a constant speed in a straight line. However, team sport movement involves continual changes in both speed and direction, both of which increase energy cost. Accordingly, new models have emerged that incorporate both speed and acceleration to determine metabolic power. This provides a more complete measure of energy expenditure in intermittent activity, and is potentially more suitable than displacement variables for research into the demands of team sports.
Article
Full-text available
Abstract Training load (TL) is monitored with the aim of making evidence-based decisions on appropriate loading schemes to reduce injuries and enhance team performance. However little is known in detail about the variables of load and methods analysis used in high level football. Therefore the aim of this study was to provide information on the practices and perceptions of monitoring in professional clubs. Eighty two high-level football clubs from Europe, the United States and Australia were invited to answer questions relating to (1) how TL is quantified; (2) how players’ responses are monitored, and (3) their perceptions of the effectiveness of monitoring. Forty one responses were received. All teams used GPS and heart rate monitors during all training sessions and 28 used RPE. The top 5 ranking TL variables were; acceleration (various thresholds), total distance, distance covered above 5.5 m·s-1,estimated metabolic power, and heart rate exertion. Players’ responses to training are monitored using questionnaires (68% of clubs) and submaximal exercise protocols (41%). Differences in expected vs. actual effectiveness of monitoring were 23% and 20% for injury prevention and performance enhancement respectively (P<0.001 d=1.0 to 1.4). Of the perceived barriers to effectiveness, “limited human resources” scored highest, followed by “coach buy-in”. The discrepancy between expected and actual effectiveness appears to be due to suboptimal integration with coaches, insufficient human resources and concerns over the reliability of assessment tools. Future approaches should critically evaluate the usefulness of current monitoring tools and explore methods of reducing the identified barriers to effectiveness.
Article
Full-text available
The aims of the current study were to examine the magnitude of between-GPS-models differences in commonly reported running-based measures in football, examine between-units variability, and assess the effect of software updates on these measures. Fifty identical-brand GPS units (15 SPI-proX and 35 SPI-proX2, 15 Hz, GPSports, Canberra, Australia) were attached to a custom-made plastic sled towed by a player performing simulated match running activities. GPS data collected during training sessions over 4 wk from 4 professional football players (N = 53 files) were also analyzed before and after 2 manufacturer-supplied software updates. There were substantial differences between the different models (eg, standardized difference for the number of acceleration >4 m/s2 = 2.1; 90% confidence limits [1.4, 2.7], with 100% chance of a true difference). Between-units variations ranged from 1% (maximal speed) to 56% (number of deceleration >4 m/s2). Some GPS units measured 2-6 times more acceleration/deceleration occurrences than others. Software updates did not substantially affect the distance covered at different speeds or peak speed reached, but 1 of the updates led to large and small decreases in the occurrence of accelerations (-1.24; -1.32, -1.15) and decelerations (-0.45; -0.48, -0.41), respectively. Practitioners are advised to apply care when comparing data collected with different models or units or when updating their software. The metrics of accelerations and decelerations show the most variability in GPS monitoring and must be interpreted cautiously.
Article
Usage of accelerometers within player tracking devices in sport to quantify load, vertical ground reaction force (vGRF) or energy expenditure is contrary to placement guidelines. This study aimed to determine whether trunk-mounted accelerometers were a valid and reliable method to estimate thoracic segment or centre of gravity (COG) acceleration or vGRF, and the whether the elasticised harness contributes to the overestimation of acceleration. Ten male amateur rugby players performed five linear running tasks per lower limb at three speeds, twice, each with a different player tracking unit. Three-dimensional data were recorded and triaxial accelerometers were attached lateral to the device on the harness and skin and both shanks. Accelerometers demonstrated poor reliability (ICC:0.0-0.67), high variability (CV%:14-33%) and change in mean (41-160%), and were not valid to estimate vertical acceleration of the COG and thoracic segment nor vGRF. Caution is advised when utilising trunk-mounted triaxial accelerometer data as it is not a valid or reliable means to estimate peak vertical acceleration for its thoracic location nor whole-body COG acceleration or vGRF during running. To improve player tracking instrument validity and reliability, a new attachment method and/or harness material(s), that reduce or eliminate extraneous acceleration during running, are urgently required.
Article
The aims of the present study were 1) to analyze the association between the time score obtained by a GPS device at a frequency of 10 Hz and by timing gates, and 2) to assess intrarater reliability and agreement of these two methods in a 20 m straight sprint test. Twelve amateur soccer players (21.08 ± 1.57 yr) gave their voluntary consent to take part in this study. Each player performed 5 repetitions of a 20 m sprint test in each of the two test sessions. The start of the sprint for the GPS device was determined by a continuous increase in speed from below (0 to 0.1 km/h). No significant relationship (r = 0.118, p > 0.05) between 20 m straight sprint times measured by the timing gates and the GPS was found. The test-retest intra-class correlation coefficient (ICC) obtained from the measurement of sprint times with the GPS device was small (r = 0.118, p > 0.05). The continuous increase in speed from below (0 to 0.1 km/h) seems to not to be a precise way to evaluate the sprint time by a GPS device in a 20 m straight sprint test.
Article
We compared the accuracy of 2 GPS systems with different sampling rates for the determination of distances covered at high-speed and metabolic power derived from a combination of running speed and acceleration. 8 participants performed 56 bouts of shuttle intermittent running wearing 2 portable GPS devices (SPI-Pro, GPS-5 Hz and MinimaxX, GPS-10 Hz). The GPS systems were compared with a radar system as a criterion measure. The variables investigated were: total distance (TD), high-speed distance (HSR>4.17 m·s(-1)), very high-speed distance (VHSR>5.56 m·s(-1)), mean power (Pmean), high metabolic power (HMP>20 W·kg(-1)) and very high metabolic power (VHMP>25 W·kg(-1)). GPS-5 Hz had low error for TD (2.8%) and Pmean (4.5%), while the errors for the other variables ranged from moderate to high (7.5-23.2%). GPS-10 Hz demonstrated a low error for TD (1.9%), HSR (4.7%), Pmean (2.4%) and HMP (4.5%), whereas the errors for VHSR (10.5%) and VHMP (6.2%) were moderate. In general, GPS accuracy increased with a higher sampling rate, but decreased with increasing speed of movement. Both systems could be used for calculating TD and Pmean, but they cannot be used interchangeably. Only GPS-10 Hz demonstrated a sufficient level of accuracy for quantifying distance covered at higher speeds or time spent at very high power.
Article
To examine the validity and inter-unit reliability of 10Hz GPS for measuring instantaneous velocity during maximal accelerations. Experimental. Two 10Hz GPS devices secured to a sliding platform mounted on a custom built monorail were towed whilst sprinting maximally over 10m. Displacement of GPS devices was measured using a laser sampling at 2000Hz, from which velocity and mean acceleration were derived. Velocity data was pooled into acceleration thresholds according to mean acceleration. Agreement between laser and GPS measures of instantaneous velocity within each acceleration threshold was examined using least squares linear regression and Bland-Altman limits of agreement (LOA). Inter-unit reliability was expressed as typical error (TE) and a Pearson correlation coefficient. Mean bias±95% LOA during accelerations of 0-0.99ms(-2) was 0.12±0.27ms(-1), decreasing to -0.40±0.67ms(-1) during accelerations >4ms(-2). Standard error of the estimate ±95% CI (SEE) increased from 0.12±0.02ms(-1) during accelerations of 0-0.99ms(-2) to 0.32±0.06ms(-1) during accelerations >4ms(-2). TE increased from 0.05±0.01 to 0.12±0.01ms(-1) during accelerations of 0-0.99ms(-2) and >4ms(-2) respectively. The validity and reliability of 10Hz GPS for the measurement of instantaneous velocity has been shown to be inversely related to acceleration. Those using 10Hz GPS should be aware that during accelerations of over 4ms(-2), accuracy is compromised.
Article
To examine the difference in distance measured by two global positioning system (GPS) units of the same model worn by the same player while performing movements common to team sports. Twenty elite Australian football players completed two trials of the straight line movement (10, 20, 40 m) at four speeds (walk, jog, stride, sprint), two trials of the changes of direction (COD) courses of two different frequencies (gradual and tight), and five trials of a team sport running simulation circuit. To assess inter-unit variability for total and high intensity running (HIR) distance measured in matches, data from eight field players were collected in three Australian Hockey League (AHL) matches during the 2009 season. Each subject wore two GPS devices (MinimaxX v2.5, Catapult, Australia) that collected position data at 5 Hz for each movement and match trial. The percentage difference ±90% confidence interval (CI) was used to determine differences between units. Differences (±90% CI) between the units ranged from 9.9 ± 4.7% to 11.9 ± 19.5% for straight line running movements and from 9.5 ± 7.2% to 10.7 ± 7.9% in the COD courses. Similar results were exhibited in the team sport circuit (11.1 ± 4.2%). Total distance (10.3 ± 6.2%) and HIR distance (10.3 ± 15.6) measured during the match play displayed similar variability. It is recommended that players wear the same GPS unit for each exercise session to reduce measurement error. The level of between-unit measurement error should be considered when comparing results from players wearing different GPS units.
Article
To assess the validity and reliability of distance data measured by global positioning system (GPS) units sampling at 1 and 5 Hz during movement patterns common to team sports. Twenty elite Australian Football players each wearing two GPS devices (MinimaxX, Catapult, Australia) completed straight line movements (10, 20, 40 m) at various speeds (walk, jog, stride, sprint), changes of direction (COD) courses of two different frequencies (gradual and tight), and a team sport running simulation circuit. Position and speed data were collected by the GPS devices at 1 and 5 Hz. Distance validity was assessed using the standard error of the estimate (±90% confidence intervals [CI]). Reliability was estimated using typical error (TE) ± 90% CI (expressed as coefficient of variation [CV]). Measurement accuracy decreased as speed of locomotion increased in both straight line and the COD courses. Difference between criterion and GPS measured distance ranged from 9.0% to 32.4%. A higher sampling rate improved validity regardless of distance and locomotion in the straight line, COD and simulated running circuit trials. The reliability improved as distance traveled increased but decreased as speed increased. Total distance over the simulated running circuit exhibited the lowest variation (CV 3.6%) while sprinting over 10 m demonstrated the highest (CV 77.2% at 1 Hz). Current GPS systems maybe limited for assessment of short, high speed straight line running and efforts involving change of direction. An increased sample rate improves validity and reliability of GPS devices.
Unpacking the black box: applications and considerations for using GPS devices in sport
  • J J Malone
  • R Lovell
  • Varley
  • Mc
Monitoring accelerations with GPS in football: time to slow down?
  • M Buchheit
  • Al Haddad
  • H Simpson
  • Bm
Global positioning system standard positioning service performance standard. Washington, DC: DoD
  • Department Of Defense
Department of Defense. Global positioning system standard positioning service performance standard. Washington, DC: DoD, 2020.