ArticlePDF Available

A New Foot-Mounted Inertial Measurement System in Soccer: Reliability and Comparison to Global Positioning Systems for Velocity Measurements During Team Sport Actions

  • PlayerMaker

Abstract and Figures

The aims of this study were to i) compare a foot-mounted inertial system (PlayerMaker™) to three commercially available Global Positioning Systems (GPS) for measurement of velocity-based metrics during team sport movements and ii) evaluate the inter-unit reliability of the PlayerMaker™. Twelve soccer players completed a soccer simulation, whilst wearing a PlayerMaker™ and three GPS (GPS#1, #2 and #3). A sub-sample (n = 7) also wore two PlayerMaker™ systems concurrently. The PlayerMaker™ measured higher (p < 0.05) total distance (518 ± 15 m) compared to GPS#1 (488 ± 15 m), GPS#2 (486 ± 15 m), and GPS#3 (501 ± 14 m). This was explained by greater (p < 0.05) distances in the 1.5-3.5 m/s zone (356 ± 24 m vs. 326 ± 26 m vs. 324 ± 18 m vs. 335 ± 24 m) and the 3.51-5.5 m/s zone (64 ± 18 m vs. 35 ± 5 vs. 43 ± 8 m vs. 41 ± 8 m) between the PlayerMaker™, GPS#1, GPS#2 and GPS#3, respectively. The PlayerMaker™ recorded higher (p < 0.05) distances while changing speed. There were no systematic differences (p > 0.05) between the two PlayerMaker™ systems. The PlayerMaker™ is reliable and records higher velocity and distances compared to GPS.
Content may be subject to copyright.
Ahead of print DOI: 10.2478/hukin-2021-0010
A new foot-mounted inertial measurement system in soccer:
reliability and comparison to global positioning systems for
velocity measurements during team sport actions
Mark Waldron1,2*, Jamie Harding3, Steve Barrett4, Adrian Gray2
1Swansea University, College of Engineering, A-STEM, Swansea, UK.
2University of New England, School of Science and Technology, NSW, Australia.
3St Mary’s University, Faculty of Sport, Health and Applied Sciences, London, UK.
4Hull City Tigers FC, East Riding of Yorkshire, UK.
* = corresponding author
Dr Mark Waldron
College of Engineering
Engineering East (A120)
Swansea University,
Phone: +44 7774004973
Conflict of interest: The authors have no conflict of interest to declare
Acknowledgments: The results of the current study were presented at the United Kingdom Strength and
Conditioning Association annual conference 2019, Milton Keynes. No financial support was received for this
The aims of this study were to i) compare a foot-mounted inertial system (PlayerMaker™) to three
commercially available Global Positioning Systems (GPS) for measurement of velocity-based metrics during
team sport movements and ii) evaluate the inter-unit reliability of the PlayerMaker™. Twelve soccer players
completed a soccer simulation, whilst wearing a PlayerMaker™ and three GPS (GPS#1, #2 and #3). A sub-
sample (n = 7) also wore two PlayerMaker™ systems concurrently. The PlayerMaker™ measured higher (p <
0.05) total distance (518 ± 15 m) compared to GPS#1 (488 ± 15 m), GPS#2 (486 ± 15 m), and GPS#3 (501 ± 14 m).
This was explained by greater (p < 0.05) distances in the 1.5-3.5 m/s zone (356 ± 24 m vs. 326 ± 26 m vs. 324 ±
18 m vs. 335 ± 24 m) and the 3.51-5.5 m/s zone (64 ± 18 m vs. 35 ± 5 vs. 43 ± 8 m vs. 41 ± 8 m) between the
PlayerMaker™, GPS#1, GPS#2 and GPS#3, respectively. The PlayerMaker™ recorded higher (p < 0.05)
distances while changing speed. There were no systematic differences (p > 0.05) between the two
PlayerMaker™ systems. The PlayerMaker™ is reliable and records higher velocity and distances compared to
Key words: soccer, motion analysis, player tracking.
Wearable sensors are routinely used to track the movements of athletes during training and
competition (Cummins et al., 2013; Mallo et al., 2015). Team sports use micro-technology, containing both
micro-electromechanical systems (MEMSs) and global positioning systems (GPSs), worn in a vest between the
player’s scapulae to track athletes’ gross movements (Barrett et al., 2016). Using GPSs, time motion analysis
data, such as total distance covered, distance in selected velocity zones, accelerations and decelerations are
often reported to describe the external load of training or competition (Cummins et al., 2013). While most
micro-technology devices used in team sports contain MEMSs, output from these sensors provides segregated
performance variables and does not contribute to the calculation of velocity-based metrics (Malone et al., 2017).
The validity and reliability of a number of commercially available systems in measuring velocity-
based metrics have been described (Akenhead et al., 2014; Buchheit et al., 2014; Coutts and Duffield, 2010;
Varley et al., 2012), thus providing an understanding of the application and limitations of GPSs in a team
sports context. A common limitation among these studies has been the validity and reliability of GPS devices
to measure high- and variable-velocity movements in smaller areas. Notably, the accuracy of rapid
accelerations (> 3 m/s2) is consistently compromised when using GPS devices of varying specifications
(Akenhead et al., 2014; Buchheit et al., 2014). Low sampling rates (Varley et al., 2012), the positioning of the
device (Barrett et al., 2014, 2016), quality of the satellite signal (Karaim and Aboelmagd, 2018) and
inconsistencies within the data processing (Buchheit et al., 2014; Varley et al., 2017), have been suggested to
contribute to the error of GPS devices.
GPS devices typically underestimate criterion measures of mean velocity or distance (Duffield et al.,
2010; Vickery et al., 2014), but possess generally acceptable internal consistency, such that typical changes in
performance can be identified. However, coefficients of variation (CV) ranging between 20-78% have been
reported for the inter-unit reliability (i.e. agreement between two devices) of the same manufacturer and
model (Coutts and Duffield, 2010; Thornton et al., 2019). As a result, individual assignment of devices among
a squad of players has been recommended, as well as limiting between-player comparisons.
Alternative measurement methods are capable of tracking identical velocity-based metrics in a team
sports context, yet adopt a different technological approach. Inertial measurement units (IMUs) provide one
example and can be fitted about the person to monitor performance (van der Kruk and Reijne, 2018). Wearable
IMUs comprise accelerometers, gyroscopes and can include a magnetometer. Measurements of raw
acceleration and angular velocity are recorded during movement to detect temporal gait events on a stride-
by-stride basis (Yang et al., 2011). Integration of these data can determine velocity and orientation of various
body parts, respectively, depending on the selected anatomical placement of the IMU, without the
requirement of the GPS. IMUs are commonly fitted to the lower-limbs (shank or foot) for the purposes of gait
analysis, with early attempts to validate linear horizontal velocity reporting errors between ~ 2 and ~ 5%
(Hausswirth et al., 2009; Yang et al., 2011). However, most IMUs have not been specifically designed to
quantify movements observed among team sports players during training or competition. Given the
acceptable criterion validity of existing IMUs (Roell et al., 2018), this limitation could be overcome by a foot-
mounted unit, assuming the underlying algorithms are suitably designed to accommodate the complexity of
team sports movements and potential drift errors (Takeda et al., 2014).
It is somewhat surprising that integrated IMUs are not more frequently used for measuring velocity-
based metrics in team sports, since they can be used in- or outdoors and have fewer potential sources of
measurement error (van der Kruk and Reijne, 2018). However, the positioning of GPS-micro-technology
devices between the scapulae limits the scope of IMU output, whereas foot-mounted IMUs are capable of
recording the inertial motion and 3D orientation of individual limbs or limb segments with high accuracy
(O'Reilly et al., 2018). For example, Zaferiou et al. (2017) identified individual gait characteristics, such as
horizontal ground reaction, foot trajectories and footfall duration, alongside horizontal velocity, of high- and
low-level performers during running agility tasks using IMUs fixed to the dorsal aspects of both feet. While
such detailed analyses would be useful in a team sports context, the capacity of foot-worn IMUs to reliably
measure the velocity of players during team sports movements is unknown. Furthermore, to date, there has
been no direct comparison of foot-mounted, sports-specific IMUs to the most commonly adopted form of
velocity measurement (GPS-micro-technology), while performing movement patterns that simulate the
demands of a particular team sport. Thus, the primary aim of this study was to compare two methods of
velocity measurement; a foot-mounted IMU that has been specifically developed for soccer performance
(PlayerMaker™) to three commercially available GPS devices for the measurement of horizontal velocity
during team sport movements. The second aim was to evaluate the inter-unit reliability of the PlayerMaker™
Participants visited the testing facility twice. After initial familiarisation (visit 1) to the protocol and
the instrumentation, participants came to the research facility on a separate day to perform 5-min and 20-s of
a soccer-specific, multi-directional intermittent movement protocol (SAFT90; Barrett et al., 2013). Participants
were fitted with a foot-worn IMU (PlayerMaker™) and three separate GPS devices, which are most commonly
used in elite team sports, but were anonymised for the current article. GPS#1 sampled at 18 Hz, GPS#2 and
GPS#3 sampled at 10 Hz. Fitting of the IMU (PlayerMaker™) and the three GPS devices facilitated a method
comparison of velocity-based performance metrics.
Twelve elite League 1 academy-level soccer players (age 15 ± 3, range 11-18 years; body mass 54.5 ±
14.9 kg) provided written informed consent and, where necessary, parental assent to participate in this study.
Institutional ethical approval was provided for this study (SMEC_2018-19_015). Participants were informed
of the benefits and risks of the investigation prior to signing the institutionally approved informed consent
document to participate in the study
The SAFT90 was designed to replicate the movement demands of English Championship-level soccer,
based on time-motion analysis of match play, dictating the actions and pace of the movement via pre-recorded
audio instructions. The SAFT90 course is 20 m in length, comprising varying intensity (stand, walk, jog, stride
and sprint) and multi-directional actions (forward, backward and sideward locomotion) (Figure 1). The
participants’ intended route through the cone-marked course is described in Figure 1, totalling a distance of
484 m. However, given the deviation from the intended route of all participants while completing the course,
this was not used to determine criterion distance. Participants were instructed to follow the audio cues in the
same way as their familiarisation trials and ensure maximal effort during the sprint sections. All trials were
performed on a flat, well-groomed, real grass pitch, under fair weather conditions, with participants wearing
studded soccer boots and a standardized training kit issued by the soccer team. Testing was performed on the
middle of the pitch to minimise the effects of the local built environment on GPS signal quality.
Instrumentation and data collection
The same three GPS devices where tightly fitted to each participant using custom-designed neoprene
vests, placing the units 2.5 cm apart, in parallel between the scapulae (Thornton et al., 2019). The GPS devices
were kept in the same position on each participant for all trials (Left unit - GPS#2; Central unit - GPS#3; Right
Unit - GPS#1). All GPS devices were activated 20-min prior to testing to ensure that the maximal number of
satellite connections were established. There was an acceptable number of identified satellites and horizontal
dilution of precision for GPS#3 (10 ± 2 and 0.8 ± 0.1, respectively), GPS#2 (16 ± 1 and 0.7 ± 0.1, respectively) and
GPS#1 (13 ± 2 and 0.4 ± 0.1, respectively). Participants were instructed to stand motionless prior to each trial
to assist with data synchronisation. All devices were synchronized post-hoc using local time, as measured on
board the GPS devices. While this was typically sufficient, the raw velocity traces were visually checked to
verify temporal similarity, using the continuous acceleration from quiet standing as an indication of the
initiating movement. All data were uploaded to each manufacturer’s software package and ‘raw’ (unfiltered
10 Hz) velocity traces were extracted into Microsoft Excel 2016 (Microsoft Inc. WA, US.). These data were
subsequently interpolated and filtered (see data analysis).
Participants were simultaneously fitted with a PlayerMaker™ system, which is a footwear sensor unit,
housed within custom silicone straps, securely fixed over each football boot, sitting on the lateral aspect of the
calcanei (Figure 2). The PlayerMaker™ was activated 10 min prior to the session and required no calibration
by the user prior to data collection. Participants were instructed to stand motionless prior to each trial to assist
with data synchronisation. The PlayerMaker™ sensor is an IMU, comprising a 3-axis 16 g accelerometer and
3-axis gyroscope (MPU-9150, InvenSense, California, USA) for measurement of accelerations and angular
velocity of each foot during gait, respectively.
The PlayerMaker™ system calculates whole-body velocity-based metrics using data generated by the
on-board MEMS, which utilizes a combination of proprietary gait tracking and foot-based event detection
algorithms. In brief, the soccer-specific gait tracking algorithm permits detection of the orientation and
translation of the participants‘ limbs during gait cycles, while the event detection algorithm identifies key
events during gait (heel strike, toe-off, zero-velocity, zero height, non-gait pattern). The microprocessor
receives accelerometer and gyroscope data, from which orientation, velocity and position vectors are
determined utilizing a Kalman Filter together with gait events update (Figure 3). The resulting output can
provide a number of metrics germane to soccer performance; however, velocity profiles were extracted for
comparison to the GPS devices.
Inter-unit reliability
In a sub-analysis, seven of the participants (age 15 ± 2 years; body mass 57.5 ± 12.5 kg) wore two
PlayerMaker™ systems concurrently (two on each foot) during the completion of the SAFT90 to facilitate inter-
unit reliability analysis. The units were fitted adjacent to one another on the lateral aspect of the calcaneus.
The same two units were worn throughout the study.
Data analysis
Raw velocity files were downloaded and exported to Microsoft Excel 2016 (Microsoft Inc. WA, US) for
further analysis. All velocity traces were interpolated to 25 Hz signals to facilitate temporal comparisons and
to remove sampling frequency differences between the devices. 25 Hz represents the down-graded sampling
rate of the PlayerMaker™ from 1000 Hz and was deemed a suitable frequency for comparative purposes,
owing to the typical centre of mass (COM) movement frequency during human locomotion ~ 10 Hz (Welk,
2002) and the minimum sampling rate determined by the Nyquist principle. Given the reported differences in
output between software-derived and raw GPS velocity traces (Thornton et al., 2019), GPS raw data from
GPS#2 and GPS#1 devices were identically filtered using a zero-lag exponential filter (Malone et al., 2017;
Varley et al., 2017). The GPS#3 uses a median filter (personal communication) for velocity data, which we chose
to maintain as this represents the output of the device in practice. The PlayerMaker™ data were down sampled
from its origin of 1000 Hz to 25 Hz using a zero-lag Butterworth filter.
A total of 19 different time motion analysis variables were selected for analysis during the SAFT90,
based on those typically reported to describe soccer performance (Vigh-Larsen et al., 2018; Waldron and
Murphy, 2013): total distance (m); mean velocity (m/s); peak velocity (m/s); distance < 1.5 m/s (m); distance
1.5-3.5 m/s (m); distance 3.51-5.5 m/s (m); distance > 5.5 m/s (m); peak acceleration (m/s2); peak deceleration
(m/s2); acceleration count < 1.5 m/s2; acceleration count 1.5-3.5 m/s2; acceleration count > 3.5 m/s2; deceleration
count < -1.5 m/s2; deceleration count -1.5 - -3.5 m/s2; acceleration count < -3.5 m/s2; mean acceleration distance
(m); total acceleration distance (m); mean deceleration distance (m); total deceleration distance (m). The period
selected for acceleration or deceleration measurements was 0.3 s (Malone et al., 2017; Varley et al., 2017), with
minor accelerations (< 0.02 m/s2) removed to reduce low-intensity acceleration or deceleration counts. Peak
velocity was averaged over the fastest 0.3-s period during the SAFT90.
Statistical analysis
Systematic biases between the PlayerMaker™ system and each of the three GPS devices were
performed for all time motion analysis variables using paired t-tests (Atkinson and Nevill, 1998), with
statistical significance accepted at p < 0.05, with post-hoc Bonferroni adjustments. No statistical comparisons of
GPS devices were necessary to conduct. The 95% Limits of Agreement (95% LoA) were used to identify the
degree of systematic bias and random error (SD of the between system differences × 1.96) (Atkinson and
Nevill, 1998) between the PlayerMaker™ system and each of the three GPS devices. Reference to ‘total error’
hereafter refers to the sum of systematic and random error. The recommended checks for heteroscedasticity
(Atkinson and Nevill, 1998) were performed on the data generated during the SAFT90, with no significant
relationships found (p > 0.05). All statistical comparisons were performed in IBM SPSS (Software V24.0, IBM,
NY, USA), with statistical significance accepted at p < 0.05.
PlayerMaker™ to GPS comparison during the Saft90 protocol
As presented in Tables 1 & 2 and Figure 4, there were differences (p < 0.05) between most
PlayerMaker™ and GPS variables during the SAFT90. However, there were no differences (p > 0.05) for peak
velocity comparisons, distance > 5.5 m/s between the GPS#1, GPS#3 and PlayerMaker. The PlayerMaker™
consistently recorded greater (p < 0.05) total distances compared to all GPS devices (mean range = 17 - 32 m
underestimation), as well as greater distance covered between 1.5-3.5 m/s (mean range = 20 – 32 m
underestimation) and 3.51-5.5 m/s (mean range = 22 – 30 m underestimation). The distance covered between
3.51-5.5 m/s produced the widest LoA with the GPS#1 device (total error of ~ 60 m). For total distance, the total
error between the GPS devices and the PlayerMaker™ ranged from ~ 25 to 50 m.
Similarly, peak accelerations (mean range = 0.3 – 1.75 m/s2 underestimation) and decelerations (mean
range = 0.6 – 1.56 m/s2 underestimation) were of smaller magnitude when measured by GPS devices compared
to PlayerMaker™. This translated to a total error of up to ~ 3.1 m/s2 between the PlayerMaker™ and the GPS
devices (Tables 1 & 2). However, the differences (mean systematic biases) were not all in the same direction,
with the GPS devices all registering a higher number of lowest intensity (< 1.5 m/s2) accelerations (mean range
= 29-130 overestimation) and decelerations (mean range = 111 and 195 overestimation for GPS#3 and GPS#1,
respectively). This trend was reversed for higher intensity accelerations. Total acceleration distance was higher
for all of the GPS devices (p < 0.05; range = 8 – 12 m), but mean acceleration distance was higher for the
PlayerMaker™ (p < 0.05; range = 0.02 – 0.08 m). Total deceleration (p < 0.05; range = 9 – 20 m) and mean
deceleration (p < 0.05; range = 0.02 – 0.17 m) distance was lower when measured by all GPS devices. Figure 5
shows the velocity profile of a representative participant during a 60-s segment of the SAFT90 to highlight the
above differences.
PlayerMaker™ inter-unit reliability during the Saft90 protocol
As presented in Tables 1 & 2, the PlayerMaker™ system was not systematically different to its
comparative unit for any variable (p > 0.05) during the SAFT90. The 95% LoA revealed that the low-intensity (<
1.5 m/s2) acceleration and deceleration counts were least reliable, with random variation of ± 21 and ± 16. The
random error of total distance between the two PlayerMaker™ systems was ± 1.28 m during the total 5-min,
20-s SAFT90 protocol. Figure 6 shows the velocity traces of two PlayerMaker™ representative systems during
a 75-s segment of the SAFT90.
The first aim of this study was to compare the PlayerMaker™ system to three commercially available
GPS devices for the measurement of velocity-based metrics during team sports movements. We found a
number of differences between methods, with measurements such as total distance covered and distance in
three velocity zones higher in the PlayerMaker™ system. These differences were accompanied by larger peak
accelerations and peak decelerations, as well as a higher number of high-intensity velocity changes, measured
by the PlayerMaker™. The GPS devices tended to measure a larger number of low-intensity accelerations and
decelerations (particularly the GPS#1 and GPS#3 devices), leading to larger total acceleration distances, yet
lower mean acceleration and deceleration distances. Despite mean velocity discrepancies during the SAFT90,
there were no differences in peak velocity between all devices. It would appear that the change in method
(technological and anatomical differences) leads to differences in the quantification of time-motion analysis
data during intermittent team sports activity, with the IMU generally recording higher distances, velocities
and velocity changes compared to the GPS devices. These differences were not apparent when the
PlayerMaker™ system’s inter-unit reliability was assessed, demonstrating the consistency of these devices for
measuring team sports movement patterns.
GPS devices have a number of well-documented error sources, including orbital error, satellite clock
error, ionospheric error, tropospheric error and multipath and receiver noise (Karaim and Aboelmagd, 2018).
The accumulation of these errors, mixed with limited raw data sampling frequency, has led to underestimation
of criterion velocity and distance covered by ~ 10-30% (Duffield et al., 2010; Vickery et al., 2014). Whist we did
not have a reference system in the current study, it is worth recognising that the GPS devices also recorded
lower values for a number of velocity-based metrics compared to the PlayerMaker™ system. Spatio-temporal
gait characteristics measured by more rudimentary IMUs typically compare closely to gold-standard 3D
camera systems (Kluge et al., 2017), providing further confidence that the MEMS included in the
PlayerMaker™ are less susceptible to technical error. However, the complexity of velocity measurement
during team sports movements provides additional challenges to the calculation of whole-body velocity when
using IMUs. Indeed, while IMUs do not have any of the same limitations as GPS devices, a problem often
experienced is so-called ‘drift error’. Here, IMUs fitted to lower-limbs can measure velocity of body segments
by integration of raw acceleration; however, this leads to cumulative signal errors across time (i.e. drift). The
PlayerMaker™ system corrects for these errors using a Kalman Filter, whereby zero-velocity updates are used
to calibrate the sensors. While these approaches are well-known for simple locomotive patterns, such as linear
walking or running (Takeda et al., 2014), they are insufficient for estimating velocity when gait patterns are
more complex. The PlayerMaker™ uses advanced gait phase detection for soccer-specific movement, based
on a custom-built machine learning algorithm. The advanced gait phase detection process detects the zero-
velocity phases more efficiently and, in addition, detects zero-height phases, which usually occur during the
stance phase of gait. The aforementioned detections are used as input to the Kalman Filter and affect the
position and velocity estimation, thus controlling drift errors. These processes reduce the error of the system
and assist with the accuracy of the PlayerMaker™.
There are obvious differences in the technology and algorithmic approaches used to calculate velocity
between GPS and IMUs, which provide some explanation for the differences observed in the current study.
However, it is noteworthy that the anatomical placement of the devices is markedly different and will
influence the data generated during team sports movement patterns (Barrett et al., 2014, 2016). Measurement
of the COM (or spinal alignment thereof) does not reflect the entire movement of the lower limbs, leading to
discrepancies in movement detection between lower limbs and the COM during soccer specific activities
(Barrett et al., 2016; Nedelec et al., 2014). The larger difference of the PlayerMaker™ to the 484 m ‘intended
route’ perhaps demonstrates this, where the players’ foot placement will inevitably deviate from the body’s
centre line and cover greater distances.
The ability to monitor team sport activities using the GPS or IMUs has been historically difficult, due
to the random, intermittent and multi-directional nature of movements. Short, rapid and forceful movements
are necessary to efficiently increase or decrease velocity while completing complex movement patterns,
typically resulting in greater horizontal displacement of the ground contact points (i.e. the feet) relative to the
COM (Morrison et al., 2015). Based on this reasoning, the GPS unit placed between the scapulae and the foot-
mounted sensors will follow separate paths during team sports running patterns. Furthermore, the
PlayerMaker™ system is placed on two anatomical landmarks and must approximate horizontal velocity
across both limbs. Given the complex nature of movements required to perform the SAFT90, it is assumed that
the PlayerMaker™ measures a number of minor foot displacements, which would not register as clearly via a
single unit placed between the scapulae. This is, perhaps, more likely among well-trained athletes, who
typically attempt to maintain a narrower base of support and reduce turning arcs during change of direction
tasks, resulting in less displacement of the upper trunk, in favour of higher frequency ground contacts
(Zaferiou et al., 2017). This anatomical placement is important to the results of the current study and, alongside
the higher sensitivity of the PlayerMaker™, is likely to explain much of the variation between systems.
Based on the above reasons, the PlayerMaker™ system quantifies rapid changes in velocity during
soccer-based movements that will be unregistered by GPS devices, owing to a change in the method of
measurement (i.e. technology and anatomical placement). The change in method leads to greater distances
being reported, particularly in higher velocity zones, alongside larger mean acceleration or deceleration
distances. The higher acceleration distance measured by all GPS units is explained by higher frequency of very
low-intensity accelerations and might reflect the inability of the GPS signal to register higher-intensity velocity
changes compared to IMUs. This is more likely because, unlike the GPS-velocity, the PlayerMaker™ system
is not limited by a sampling rate, with a raw sampling rate of 1000 Hz. Thus, fast directional change with high
foot cadence can be readily tracked. These findings are of critical importance to soccer practitioners, as high-
intensity actions are associated with elite soccer performance levels (Waldron and Murphy, 2013) and
differentiate between positions (Vigh-Larsen et al., 2018). Moreover, the PlayerMaker™ registered greater
deceleration distances and high-intensity counts, which is consistent with the poorer validity of 10 Hz GPS
devices in measuring decelerations (11.3% CV), which are typically of larger magnitude compared to
accelerations (Varley et al., 2012). This is important, since decelerations are associated with increased muscle
soreness and impaired neuromuscular function following soccer matches (Nedelec et al., 2014) and, therefore,
require accurate quantification.
Among all of these measures, there were no systematic inter-unit differences for the PlayerMaker™
and the largest 95% LoA for distance covered being ~ 4.5 m (Table 2) across all velocity zones. Small total errors
in distance travelled would permit detection of a number of differences, including differences in match-
running performance of elite and sub-elite soccer players (Waldron and Murphy, 2013). There was, however,
a larger random variation for total acceleration distance (± 12.52 m; Table 2) between units, indicating slightly
larger noise for this variable that might preclude detection of more refined signal changes. However, this
random noise would still enable detection of changes in acceleration distance that might be important in
practice. For example, acceleration distance changes by a total of ~ 50 m across intensity zones between the
first and the second 15-min period of professional matches (Akenhead et al., 2014), which is much larger than
the inherent noise. Similarly, despite the systematic differences between methods and total error of up to ~ 25
m (GPS#1, Table 1), this noise is still within a ~ 50 m signal change and would, therefore, not significantly alter
the interpretation between methods. The disagreement between methods and random error within the
PlayerMaker™ would not become more noticeable until smaller changes in acceleration distance, such as the
5-10 m declines between final match periods, are necessary to quantify (Akenhead et al., 2014). Of interest, the
random error of deceleration distances was lower (~ 6 m; Table 2) for the PlayerMaker™, which we speculate
might be related to the motion of the ankle during these movements compared to accelerations. This is beyond
the scope of the current study, but is worthy of further investigation.
Soccer researchers and practitioners should, therefore, consider the current results in relation to their
desired outcomes. That is, the PlayerMaker™ system is more consistent between units compared to previous
GPS reports (Coutts and Duffield, 2010; Thornton et al., 2019) and will bias its measurement of whole-body
velocity towards the movement of the lower-limbs. This is arguably of greater importance to soccer
practitioners, since understanding of work done by the lower limbs during team sports movement patterns
has been incorporated into recent mechanical energetic models (Gray et al., 2018) and it is lower-limb measures
of muscle function that are often prioritised to determine exercise-induced fatigue in practice (McCall et al.,
2015). Furthermore, foot mounted IMUs have the capacity to measure an array of mechanical loading/gait
parameters, which have not been explored in the current study, but could be used by practitioners to assess
performance, asymmetries and neuromuscular function of players during training or competition.
Soccer practitioners should be aware of the differences between these two distinctly different methods
(GPS or foot-worn IMU) for the measurement of over-ground velocity. Distances covered in higher velocity
zones, peak accelerations and decelerations and high-intensity velocity changes are higher when measured
using foot-mounted IMUs (PlayerMaker™) compared to three commercially available GPS devices. Mean
velocity, but not peak velocity, also differs between these two types of technology. Practitioners can, therefore,
use foot-worn IMUs or GPS devices for tracking players during training and competition (depending on
governing body regulations), but understand that movement of the lower-limbs during short and rapid
changes will be directly incorporated into the PlayerMaker™ velocity measurement, while GPS velocity will
be based on displacements of a single sensor placed superior to the COM. If it is desirable for practitioners to
bias this measurement toward lower-limb movements, then the foot-worn system would be preferable. Those
working with soccer players should also consider that the PlayerMaker™ has inter-unit reliability that would
enable interchangeable interpretations for almost all time motion data. This would be useful for comparisons
between players, without concern over noise emanating from technical errors. Further research on the
PlayerMaker™ system to soccer performance will help understand its potential applications more thoroughly.
Akenhead R, French D, Thompson KG, Hayes PR. The acceleration dependent validity and reliability of 10 Hz
GPS. J Sci Med in Sport, 2014; 17: 562–566
Atkinson G, Nevill AM. Statistical methods for addressing measurement error (reliability) in variables relevant
to sports medicine. Sport Med, 1998; 26: 217-238
Barrett S, Guard A, Lovell R. SAFT90 simulates the internal and external loads of competitive soccer match-
play. In Science and Football VII: Proceedings of the Seventh World Congress on Science and Football,
Chapter: 15, Publisher: Routledge, Editors: H. Nunome, B. Drust and B. Dawson, pp.95-100, 2013
Barrett S, Midgley A, Lovell R. PlayerLoad™: reliability, convergent validity and influence of unit position
during treadmill running. Int J Sport Physiol Perform, 2014; 9: 945-952
Barrett S, Midgley AW, Towlson C, Garrett A, Portas M, Lovell R. Within-match Playerload™ patterns during
a simulated soccer match: potential implications for unit positioning and fatigue management. Int J
Sport Physiol Perform, 2016; 11: 135-140
Buchheit M, Al Haddad H, Simpson BM, Palazzi D, Bourdon PC, Di Salvo V, Mendez-Villanueva A.
Monitoring accelerations with GPS in football: time to slow down? Int J Sports Physiol Perform, 2014; 9:
Coutts A, Duffield R. Validity and reliability of GPS devices for measuring movement requirements in team
sports. J Sci Med Sport, 2010; 13:133-135
Cummins C, Orr R, O’Connor H, West C. Global positioning systems (GPS) and microtechnology sensors in
team sports: a systematic review. Sport Med, 2013; 43: 1025-1042
Duffield R, Reid M, Baker J, Spratford W. Accuracy and reliability of GPS devices for measurement of
movement patterns in confined spaces for court-based sports. J Sci Med Sport, 2010; 15: 523-525
Gray AJ, Shorter K, Cummins C, Murphy A, Waldron M. Modelling movement energetics using global
positioning system devices in contact team sports: limitations and solutions. Sports Med, 2018; 48: 1357-
Hausswirth C, Le Meur Y, Couturier A, Bernard T, Brisswalter J. Accuracy and repeatability of the Polar1
RS800sd to evaluate stride rate and running speed. Int J Sport Med, 2009, 30: 354–359
Karaim ME, Aboelmagd N. GNSS Error sources, multifunctional operation and application of GPS. In: Rustam
B. Rustamov and Arif M. Hashimov, IntechOpen, 2018: DOI: 10.5772/intechopen.75493
Kluge F, Gaßner H, Hannink J, Pasluosta C, Klucken J, Eskofierm BM. Towards mobile gait analysis:
concurrent validity and test-retest reliability of an inertial measurement system for the assessment of
spatio-temporal gait parameters. Sensors, 2017; 17:1522. doi: 10.3390/s17071522
Malone JJ, Lovell R, Varley MC, Coutts AJ. Unpacking the black box: applications and considerations for using
GPS devices in sport. Int J Sports Physiol Perform, 2017; 12: 18-26
Mallo J, Mena E, Nevado F, Paredes V. Physical Demands of Top-Class Soccer Friendly Matches in Relation to
a Playing Position Using Global Positioning System Technology. J Hum Kinet, 2015, 47: 179-188.
McCall A, Nedelec M, Carling C, Le Gall F, Berthoin S, Dupont G. Reliability and sensitivity of a simple
isometric posterior lower limb muscle test in professional football players. J Sports Sci, 2015; 33: 1298-
Morrison K, Albert WJ, Kuruganti U. Biomechanical assessment of change of direction performance in male
university soccer players. 33rd International Conference on Biomechanics in Sports, Poitiers, France, June 29
- July 3, 2015
Nedelec M, McCall A, Carling C, Legall F, Berthoin S, Dupont G. The influence of soccer playing actions on
the recovery kinetics after a soccer match. J Strength Cond Res, 2014; 28: 1517–1523
O'Reilly M, Caulfield B, Ward T, Johnston W, Doherty C. Wearable Inertial Sensor systems for lower limb
exercise detection and evaluation: a systematic review. Sports Med, 2018; 48: 1221-1246
Roell M, Roecker K, Gehring D, Mahler H, Gollhofer A. Player monitoring in indoor team sports: concurrent
validity of inertial measurement units to quantify average and peak acceleration values. Front Physiol,
2018; 27: 141. doi: 10.3389/fphys.2018.00141
Takeda R, Lisco G, Fujisawa T, Gastaldi L, Tohyama H, Tadano S. Drift removal for improving the accuracy
of gait parameters using wearable sensor systems. Sensors, 2014; 14: 23230-47. doi: 10.3390/s141223230
Thornton HR, Nelson AR, Delaney JA, Serpiello FR, Duthie GM. Interunit reliability and effect of data-
processing methods of global positioning systems. Int J Sports Physiol Perform, 2019; 14: 432-438
van der Kruk E, Reijne MM. Accuracy of human motion capture systems for sport applications; state-of-the-
art review. Eur J Sport Sci, 2018; 18: 806-819
Varley MC, Fairweather I, Aughey, RJ. Validity and reliability of GPS for measuring instantaneous velocity
during acceleration, deceleration and constant motion. J Sport Sci, 2012; 30: 121-127
Varley MC, Jaspers A, Helsen WF, Malone JJ. Methodological considerations when quantifying high-intensity
efforts in team sport using Global Positioning System technology. Int J Sports Physiol Perform, 2017;
Vickery WM, Dascombe BJ, Baker JD, Higham DG, Spratford WA, Duffield R. Accuracy and reliability of GPS
devices for measurement of sports-specific movement patterns related to cricket, tennisand field-based
team sports. J Strength Cond Res, 2014; 28: 1697–1705
Vigh-Larsen JF, Dalgas Uandersen TB. Position-specific acceleration and deceleration profiles in elite youth
and senior soccer players. J Strength Cond Res, 2018; 32: 1114-1122
Waldron M, Murphy AA. comparison of physical abilities and match performance characteristics among elite
and sub-elite under-14 soccer players. Pediatr Exerc Sci, 2013; 25: 423-34
Welk G. Use of accelerometry-base activity monitors to assess physical activity. In: Welk G, editor. Physical
Activity Assessments for Health-Related Research. Champaign, IL: Human Kinetics; 2002. pp. 125–141
Yang S, Mohr C, Li Q. Ambulatory running speed estimation using an inertial sensor. Gait Posture, 2011; 34:
Zaferiou AM, Ojeda L, Cain SM, Vitali RV, Davidson SP, Stirling L, Perkins NC. Quantifying performance on
an outdoor agility drill using foot-mounted inertial measurement units. PLoS One, 2017; 12, e0188184:
doi: 10.1371/journal.pone.0188184
Figure 1. The SAFT
course performed by participants. The order of events is: Blue (forwards, backwards or
sidesteps), Red (Accelerate), Green (sidestep right-left), Red (accelerate, 180° turn, sprint). The course was
repeated 11 times at random and intermittent velocities.
Figure 2. The PlayerMaker™ system fitted to the ankle of the right foot. The complete system includes
an identically sized and fitted inertial measurement unit for the left ankle.
Figure 3. PlayerMaker™ gait tracking flowchart. 502 - accelerometer, 503 - gyroscope, 504 – "R" rotation matrix
(sensor relative to local frame), 506, 508, 510 – numeric integration of raw inputs, 512 - proprietery machine
learning for gait phase detection, 514 - proprietary Kalman Filter design to calculate position, velocity and
orientation. "+" sign indicates a sum, the "X" indicates a cross product. The rotation matrix transforms the
accelerations from the sensor frame to the local frame. At block 502, the processor receives sensor data from
the accelerometer. At block 503, the processor receives the sensor data from the gyroscope. At block 504 the
acceleration data rotated to the local frame and then subtracted by g on the local z-axis. At block 506, the
processed acceleration data are integrated and the velocity vector is formed at block 508. The velocity is
integrated and the position vector, along with the velocity vector, is used at block 514, with the Kalman Filter.
At block 510 the gyroscope data are integrated for calculation of R (that is used in block 504) and for the
detection of zero-velocity update and stance, along with the raw acceleration and gyroscope data, at block 512.
(Solid lines = feedforward to gait phase detection or Kalman Filter; dashed lines = feedback)
Figure 4. Distance covered in velocity zones of the PlayerMaker™ and three Global Positioning System (GPS)
devices. * = significantly (p < 0.05) different to all GPS devices. ƚ = significantly (p < 0.05) different to
Figure 5. Velocity profiles of the PlayerMaker™ and three GPS devices 60-s segment of the SAFT
protocol in a representative participant. COD = change of direction.
Figure 6. Velocity traces of two PlayerMaker™ systems during a 75-s segment of the SAFT
in a representative participant.
Table 1. Mean differences and 95% Limits of Agreement (95% LoA) for time motion analysis
variables between the PlayerMaker™ (inertial measurement unit) and Global Positioning System
#1 and #2 (GPS) devices during the SAFT90 protocol.
Movement variables PlayerMaker™ vs.
PlayerMaker™ vs.
Mean velocity (m/s) -0.17 ± 0.12* -0.09 ± 0.05*
Peak velocity (m/s) 0.03 ± 0.71 0.12 ± 0.41
Total distance (m) -30.08 ± 12.50* -32.15 ± 19.00*
Distance < 1.5 m/s (m) 29.14 ± 17.26* 18.99 ± 17.28*
Distance 1.5-3.5 m/s (m) -30.05 ± 32.45* -31.85 ± 33.05*
Distance 3.51-5.5 m/s (m) -29.06 ± 29.41* -21.39 ± 25.41*
Distance > 5.5 m/s (m) -0.11 ± 2.53 2.10 ± 4.51*
Peak acceleration (m/s2) -1.55 ± 1.61* -0.34 ± 1.02*
Acceleration count < 1.5 m/s2 49 ± 40* 29 ± 31*
Acceleration count 1.5-3.5 m/s2 -21 ± 12* 7 ± 15*
Acceleration count > 3.5 m/s2 -6 ± 5* -4 ± 6*
Peak deceleration (m/s2) -1.56 ± 1.38* -0.60 ± 1.33*
Deceleration count < 1.5 m/s2 195 ± 69* -26 ± 44*
Deceleration count 1.5-3.5 m/s2 -22 ± 17* 6 ± 14*
Deceleration count > 3.5 m/s2 -2 ± 3* -1 ± 3*
Mean acceleration distance (m) -0.02 ± 0.04* -0.02 ± 0.03*
Total acceleration distance (m) 10.92 ± 15.89* 8.22 ± 17.35*
Mean deceleration distance (m) -0.17 ± 0.04* -0.02 ± 0.03*
Total deceleration distance (m) -20.47 ± 11.83* -20.20 ± 17.44*
Note: * = significantly different (p < 0.05) to the corresponding PlayerMaker™ system during the
SAFT90. Minus value = lower than PlayerMaker™
Table 2. Mean differences and 95% Limits of Agreement (95% LoA) for time motion
analysis variables within the PlayerMaker™ (inertial measurement unit) and
compared to Global Positioning System #3 (GPS) device during the SAFT
Movement variables PlayerMaker™ vs.
PlayerMaker™ vs.
Mean velocity (m/s) -0.05 ± 0.03* 0.00 ± 0.00
Peak velocity (m/s) -0.09 ± 0.40 0.06 ± 0.25
Total distance (m) -16.78 ± 9.45* -0.16 ± 1.28
Distance < 1.5 m/s (m) 27.77 ± 7.16* 1.22 ± 2.79
Distance 1.5-3.5 m/s (m) -20.90 ± 26.47* -1.58 ± 4.46
Distance 3.51-5.5 m/s (m) -23.31 ± 27.15* 0.02 ± 2.38
Distance > 5.5 m/s (m) -0.35 ± 2.58 0.18 ± 0.81
Peak acceleration (m/s2) -1.75 ± 0.91* 0.04 ± 0.77
Acceleration count < 1.5 m/s2 130 ± 54* -3 ± 21
Acceleration count 1.5-3.5 m/s2 -29 ± 13* -1 ± 6
Acceleration count > 3.5 m/s2 -6 ± 5* 0 ± 1
Peak deceleration (m/s2) -1.56 ± 1.30* -0.09 ± 0.61
Deceleration count < 1.5 m/s2 111 ± 68* -1 ± 16
Deceleration count 1.5-3.5 m/s2 -27 ± 10* -3 ± 3
Deceleration count > 3.5 m/s2 -2 ± 3* 0 ± 1
Mean acceleration distance (m) -0.08 ± 0.03* 0.00 ± 0.01
Total acceleration distance (m) 11.77 ± 13.01* -0.56 ± 12.52
Mean deceleration distance (m) -0.09 ± 0.05* 0.00 ± 0.01
Total deceleration distance (m) -8.58 ± 9.42* -2.13 ± 5.98
Note: * = significantly different (p < 0.05) to the corresponding PlayerMaker™
system during the SAFT90. Minus value = lower than PlayerMaker™
... For the inertial measuring device (IMU), EEE calculations were done by the manufacturer as the software lacked this feature. The technical properties of this device is explained elsewhere [28] and it applies the same metabolic power method [16]. Specifically, speed and acceleration were calculated in 10 Hz, together with the formula and constants provided in the algorithm by [16]. ...
... Nonetheless, acceptable validity and reliability for GPS devices of 10 and 18 Hz, as used in this study, have been reported [38]. Further, the inertial sensor used have been compared against high sampling GPS units [28]. Although SD varied between the specific devices, this alone is unlikely to explain the discrepancy in EEE. ...
Full-text available
The purpose of the study was to assess the accuracy of commonly used GPS/accelerometer-based tracking devices in the estimation of exercise energy expenditure (EEE) during high-intensity intermittent exercise. A total of 13 female soccer players competing at the highest level in Norway (age 20.5 ± 4.3 years; height 168.4 ± 5.1 cm; weight 64.1 ± 5.3 kg; fat free mass 49.7 ± 4.2 kg) completed a single visit test protocol on an artificial grass surface. The test course consisted of walking, jogging, high-speed running, and sprinting, mimicking the physical requirements in soccer. Three commonly used tracking devices were compared against indirect calorimetry as the criterion measure to determine their accuracy in estimating the total energy expenditure. The anaerobic energy consumption (i.e., excess post-exercise oxygen consumption, EPOC) and resting time were examined as adjustment factors possibly improving accuracy. All three devices significantly underestimated the total energy consumption, as compared to the criterion measure (p = 0.022, p = 0.002, p = 0.017; absolute ICC = 0.39, 0.24 and 0.30, respectively), and showed a systematic pattern with increasing underestimation for higher energy consumption. Excluding EPOC from EEE reduced the bias substantially (all p’s becoming non-significant; absolute ICC = 0.49, 0.54 and 0.49, respectively); however, bias was still present for all tracking devices. All GPS trackers were biased by showing a general tendency to underestimate the exercise energy consumption during high intensity intermittent exercising, which in addition showed a systematic pattern by over- or underestimation during lower or higher exercising intensity. Adjusting for EPOC reduced the bias and provided a more acceptable accuracy. For a more correct EEE estimation further calibration of these devices by the manufacturers is strongly advised by possibly addressing biases caused by EPOC.
... On an individual level, there was lower within-subject variation across PL and PL/min. Some of this variability may be extenuated to the placement of the device (between the scapulae), with suggestions that foot-mounted inertial sensors may be a more appropriate method to capture these specific movements [44]. Still, PL has been linked to alterations in acute fatigue within a fixed soccer simulation, postulating this may be able to detect alterations within an individual's locomotor efficiency [29]. ...
Full-text available
The aim of this study was to assess the measurement properties of external training load measures across three formats of standardised training games. Eighty-eight players from two English professional soccer clubs participated in the study spanning three consecutive seasons. External training load data was collected from three types of standardised game format drills (11v11, 10v10, 7v7+6) using Global Positioning Systems. For each external training load metric in each game format, the following measurement properties were calculated; coefficient of variation (CV%) to determine between- and within-subject reliability, intraclass coefficient correlation (ICC) to determine test-retest reliability, and signal-to-noise ratio (SNR) to determine sensitivity. Total distance (TD) and PlayerLoad ™ (PL) demonstrated good sensitivity (TD SNR = 1.6–4.6; PL SNR = 1.2–4.3) on a group level. However, a wide variety of within-subject reliability was demonstrated for these variables (TD CV% = 1.7–36.3%; PL CV% = 4.3–39.5%) and corresponding intensity measures calculated per minute. The percentage contribution of individual planes to PL showed the lowest between-subject CV% (CV% = 2–7%), although sensitivity varied across formats (SNR = 0.3–1.4). High speed running demonstrated poor reliability across all three formats of SSG (CV% = 51–103%, ICC = 0.03–0.53). Given the measurement properties of external training load measures observed in this study, specifically the within-subject variation, reliability across trials of standardised training games should be calculated on an individual level. This will allow practitioners to detect worthwhile changes across trials of standardised game format drills. Such information is important for the appropriate implementation of training and monitoring strategies in soccer.
Objectives: To (i) quantify the differences in locomotor and technical characteristics between different drill categories in female soccer and (ii) explore the training drill distributions between different standards of competition. Methods: : Technical (ball touches, ball releases, high-speed ball releases) and locomotor data (total distance, high-speed running distance [>5.29 m∙s-1]) were collected using foot mounted inertial measurement units from 458 female soccer players from three Women's Super League (WSL; n = 76 players), eight Women's Championship (WC; n = 217) and eight WSL Academy (WSLA; n = 165) teams over a 28-week period. Data was analysed using general linear mixed effects. Results: Across all standards, the largest proportion of time was spent in technical (TEC) (WSL = 38%, WC = 28%, WSLA = 29%) and small sided extensive games (SSGe) (WSL = 20%, WC = 31%, WSLA = 30%) drills. WSL completed more TEC and tactical (TAC) training whilst WC and WSLA players completed more SSGe and possession (POS) drills. Technical drills elicited the highest number of touches, releases and the highest total distance and high-speed activity. Position specific drills elicited the lowest number of touches and releases and the lowest total distance. When the technical and locomotor demand of each drill were made relative to time, there were limited differences between drills, suggesting drill duration was the main moderating factor. Conclusion: Findings provide novel understanding of the technical and locomotor demands of different drill categories in female soccer. These results can be used by coaches and practitioners to inform training session design.
Full-text available
Objectives: This study aimed to (i) establish the concurrent validity and intra-unit reliability of a foot-mounted inertial measurement unit for monitoring soccer technical actions, (ii) quantify the within-microcycle inter-positional differences in the technical actions of professional soccer training, and (iii) determine the influence of drill category on the technical actions of professional soccer training. Methods: Twenty-one professional soccer players' technical performance data (ball touches, releases, ball touches per minute, releases per minute), collected during training sessions throughout twenty-four weekly microcycles, were analysed using general linear modelling. Results: The inertial measurement unit exhibited good concurrent validity (PA = 95.1% - 100.0%) and intra-unit reliability (PA = 95.9% - 96.9%, CV = 1.4% - 2.9%) when compared with retrospective video analyses. The most ball touches (X‾ = 218.0) and releases (X‾ = 110.8) were observed on MD - 1, with MD - 5 eliciting the highest frequency of ball touches (X‾ = 3.8) and releases (X‾ = 1.7) per minute. Central midfielders performed the most ball touches (X‾ = 221.9), releases (X‾ = 108.3), ball touches per minute (X‾ = 3.4) and releases per minute (X‾ = 1.6). Small-sided games evoked more ball touches (X‾diff = 1.5) and releases per minute (X‾diff = 0.1) than previously reported in match-play. The fewest ball touches (X‾ = 1.2) and releases per minute (X‾ = 0.5) were observed during tactical drills. Conclusion: The results of this study provide a novel understanding of the within-microcycle, inter-positional and drill category differences in the technical actions performed by professional players during training.
Full-text available
The aim of this study was to examine the physical demands imposed on professional soccer players during 11-a-side friendly matches in relation to their playing position, using global positioning system (GPS) technology. One hundred and eleven match performances of a Spanish "La Liga" team during the 2010-11 and 2011-12 pre-seasons were selected for analysis. The activities of the players were monitored using GPS technology with a sampling frequency of 1 Hz. Total distance covered, distance in different speed categories, accelerations, and heart rate responses were analyzed in relation to five different playing positions: central defenders (n=23), full-backs (n=20), central midfielders (n=22), wide midfielders (n=26), and forwards (n=20). Distance covered during a match averaged 10.8 km, with wide and central midfielders covering the greatest total distance. Specifically, wide midfielders covered the greatest distances by very high-intensity running (>19.8 km·h-1) and central midfielders by jogging and running (7.2-19.7 km·h-1). On the other hand, central defenders covered the least total distance and at high intensity, although carried out more (p<0.05-0.01) accelerations than forwards, wide midfielders, and fullbacks. The work rate profile of the players obtained with the GPS was very similar to that obtained with semi-automatic image technologies. However, when comparing results from this study with data available in the literature, important differences were detected in the amount of distance covered by sprinting, which suggests that caution should be taken when comparing data obtained with the GPS with other motion analysis systems, especially regarding high-intensity activities.
Full-text available
Purpose: To establish the inter-unit reliability of a range of Global Positioning System (GPS)-derived movement indicators, determine the variation between manufacturers, and investigate the difference between software-derived and raw data. Methods: A range of movement variables were obtained from 27 GPS units from three manufacturers (GPSports EVO; 10 Hz, n = 10: STATSports Apex; 10 Hz, n = 10: and Catapult S5; 10 Hz, n = 7) that measured the same team-sports simulation session while positioned on a sled. The inter-unit reliability was determined using the coefficient of variation (CV; %) and 90% confidence limits (CL), whereas between manufacturer comparisons, and also comparisons of software versus raw processed data were established using standardized effect sizes (ES) and 90% CL. Results: The inter-unit reliability for both software and raw processed data ranged from good to poor (CV = 0.2%; ±1.5% to 78.2%; ±1.5%), with distance, speed, and maximal speed exhibiting the best reliability. There were substantial differences between manufacturers, particularly for threshold-based acceleration and deceleration variables (ES; ±90% CL [-2.0%; ±0.1 to 1.9%; ±0.1%]), and there were substantial differences between data processing methods for a range of movement indicators. Conclusions: The inter-unit reliability of most movement indicators were deemed as good regardless of processing method, suggesting that practitioners can have confidence within systems. Standardized data processing methods are recommended, due to the large differences between data outputs from various manufacturer-derived software.
Full-text available
Objective: Sport research often requires human motion capture of an athlete. It can, however, be labour-intensive and difficult to select the right system, while manufacturers report on specifications which are determined in set-ups that largely differ from sport research in terms of volume, environment and motion. The aim of this review is to assist researchers in the selection of a suitable motion capture system for their experimental set-up for sport applications. An open online platform is initiated, to support (sport)researchers in the selection of a system and to enable them to contribute and update the overview. Design: systematic review; Method: Electronic searches in Scopus, Web of Science and Google Scholar were performed, and the reference lists of the screened articles were scrutinised to determine human motion capture systems used in academically published studies on sport analysis. Results: An overview of 17 human motion capture systems is provided, reporting the general specifications given by the manufacturer (weight and size of the sensors, maximum capture volume, environmental feasibilities), and calibration specifications as determined in peer-reviewed studies. The accuracy of each system is plotted against the measurement range. Conclusion: The overview and chart can assist researchers in the selection of a suitable measurement system. To increase the robustness of the database and to keep up with technological developments, we encourage researchers to perform an accuracy test prior to their experiment and to add to the chart and the system overview (online, open access).
Full-text available
Quantifying the training and competition loads of players in contact team sports can be performed in a variety of ways, including kinematic, perceptual, heart rate or biochemical monitoring methods. Whilst these approaches provide data relevant for team sports practitioners and athletes, their application to a contact team sport setting can sometimes be challenging or illogical. Furthermore, these methods can generate large fragmented datasets, do not provide a single global measure of training load and cannot adequately quantify all key elements of performance in contact team sports. A previous attempt to address these limitations via the estimation of metabolic energy demand (global energy measurement) has been criticised for its inability to fully quantify the energetic costs of team sports, particularly during collisions. This is despite the seemingly unintentional misapplication of the model’s principles to settings outside of its intended use. There are other hindrances to the application of such models, which are discussed herein, such as the data-handling procedures of Global Position System manufacturers and the unrealistic expectations of end users. Nevertheless, we propose an alternative energetic approach, based on Global Positioning System-derived data, to improve the assessment of mechanical load in contact team sports. We present a framework for the estimation of mechanical work performed during locomotor and contact events with the capacity to globally quantify the work done during training and matches.
Full-text available
The increasing interest in assessing physical demands in team sports has led to the development of multiple sports related monitoring systems. Due to technical limitations, these systems primarily could be applied to outdoor sports, whereas an equivalent indoor locomotion analysis is not established yet. Technological development of inertial measurement units (IMU) broadens the possibilities for player monitoring and enables the quantification of locomotor movements in indoor environments. The aim of the current study was to validate an IMU measuring by determining average and peak human acceleration under indoor conditions in team sport specific movements. Data of a single wearable tracking device including an IMU (Optimeye S5, Catapult Sports, Melbourne, Australia) were compared to the results of a 3D motion analysis (MA) system (Vicon Motion Systems, Oxford, UK) during selected standardized movement simulations in an indoor laboratory (n = 56). A low-pass filtering method for gravity correction (LF) and two sensor fusion algorithms for orientation estimation [Complementary Filter (CF), Kalman-Filter (KF)] were implemented and compared with MA system data. Significant differences (p < 0.05) were found between LF and MA data but not between sensor fusion algorithms and MA. Higher precision and lower relative errors were found for CF (RMSE = 0.05; CV = 2.6%) and KF (RMSE = 0.15; CV = 3.8%) both compared to the LF method (RMSE = 1.14; CV = 47.6%) regarding the magnitude of the resulting vector and strongly emphasize the implementation of orientation estimation to accurately describe human acceleration. Comparing both sensor fusion algorithms, CF revealed slightly lower errors than KF and additionally provided valuable information about positive and negative acceleration values in all three movement planes with moderate to good validity (CV = 3.9-17.8%). Compared to x-and y-axis superior results were found for the z-axis. These findings demonstrate that IMU-based wearable tracking devices can Roell et al. Inertial Sensors for Player Monitoring successfully be applied for athlete monitoring in indoor team sports and provide potential to accurately quantify accelerations and decelerations in all three orthogonal axes with acceptable validity. An increase in accuracy taking magnetometers in account should be specifically pursued by future research.
Full-text available
Background: Analysis of lower limb exercises is traditionally completed with four distinct methods (i) 3D motion capture; (ii) depth-camera based systems (iii) visual analysis from a qualified exercise professional; (iv) self-assessment. Each method is associated with a number of limitations. Objective: The aim of this systematic review is to synthesize and evaluate studies which have investigated the capacity for inertial measurement unit (IMU) technologies to assess movement quality in lower limb exercises. Data Sources: A systematic review of PubMed, ScienceDirect and Scopus was conducted. Study Eligibility Criteria: Articles written in English and published in the last 10 years which contained an IMU system for the analysis of repetition-based targeted lower limb exercises were included. Study Appraisal and Synthesis Methods: The quality of included studies was measured using an adapted version of the STROBE assessment criteria for cross-sectional studies. The studies were categorised in to three groupings: exercise detection, movement classification or measurement validation. Each study was then qualitatively summarised. Results: From the 2452 articles that were identified with the search strategies, 47 papers are included in this review. Conclusions: Wearable inertial sensor systems for analysing lower limb exercises are a rapidly growing technology. Research over the past ten years has predominantly focused on validating measurements that the systems produce and classifying users’ exercise quality. There have been very few user evaluation studies and no clinical trials in this field to date.
Full-text available
Running agility is required for many sports and other physical tasks that demand rapid changes in body direction. Quantifying agility skill remains a challenge because measuring rapid changes of direction and quantifying agility skill from those measurements are difficult to do in ways that replicate real task/game play situations. The objectives of this study were to define and to measure agility performance for a (five-cone) agility drill used within a military obstacle course using data harvested from two foot-mounted inertial measurement units (IMUs). Thirty-two recreational athletes ran an agility drill while wearing two IMUs secured to the tops of their athletic shoes. The recorded acceleration and angular rates yield estimates of the trajectories, velocities and accelerations of both feet as well as an estimate of the horizontal velocity of the body mass center. Four agility performance metrics were proposed and studied including: 1) agility drill time, 2) horizontal body speed, 3) foot trajectory turning radius, and 4) tangential body acceleration. Additionally, the average horizontal ground reaction during each footfall was estimated. We hypothesized that shorter agility drill performance time would be observed with small turning radii and large tangential acceleration ranges and body speeds. Kruskal-Wallis and mean rank post-hoc statistical analyses revealed that shorter agility drill performance times were observed with smaller turning radii and larger tangential acceleration ranges and body speeds, as hypothesized. Moreover, measurements revealed the strategies that distinguish high versus low performers. Relative to low performers, high performers used sharper turns, larger changes in body speed (larger tangential acceleration ranges), and shorter duration footfalls that generated larger horizontal ground reactions during the turn phases. Overall, this study advances the use of foot-mounted IMUs to quantify agility performance in contextually-relevant settings (e.g., field of play, training facilities, obstacle courses, etc.).
Full-text available
The purpose of this study was to assess the concurrent validity and test–retest reliability of a sensor-based gait analysis system. Eleven healthy subjects and four Parkinson’s disease (PD) patients were asked to complete gait tasks whilst wearing two inertial measurement units at their feet. The extracted spatio-temporal parameters of 1166 strides were compared to those extracted from a reference camera-based motion capture system concerning concurrent validity. Test–retest reliability was assessed for five healthy subjects at three different days in a two week period. The two systems were highly correlated for all gait parameters (r > 0.93). The bias for stride time was 0 ± 16 ms and for stride length was 1.4 ± 6.7 cm. No systematic range dependent errors were observed and no significant changes existed between healthy subjects and PD patients. Test-retest reliability was excellent for all parameters (intraclass correlation (ICC) > 0.81) except for gait velocity (ICC > 0.55). The sensor-based system was able to accurately capture spatio-temporal gait parameters as compared to the reference camera-based system for normal and impaired gait. The system’s high retest reliability renders the use in recurrent clinical measurements and in long-term applications feasible.
The purpose of the study was to characterize and compare the position specific activity profiles of young and senior elite soccer players with special emphasis put on accelerations and decelerations. Eight professional senior matches were tracked using the ZXY tracking system and analyzed for number of accelerations and decelerations and running distances within different speed zones. Likewise, four U19- and five U17 matches were analyzed for comparison between youth and senior players. In senior players the total distance (TD) was 10776±107 m with 668±28 and 143±10 m being high-intensity running (HIR) and sprinting, respectively. Number of accelerations and decelerations were 81±2 and 84±3, with central defenders performing the lowest- and wide players the highest number. Declines were found between first and second halves for accelerations and decelerations (11±3 %), HIR (6±4 %) and TD (5±1 %), whereas sprinting distance did not differ. U19 players performed a higher number of accelerations, decelerations and TD compared to senior players. In conclusion, differences in number and distribution of accelerations and decelerations appeared between player positions, which is of importance when monitoring training and match loads and when prescribing specific training exercises. Further, youth players performed as much high-intensity activities as senior players indicating that this is not a discriminating physiological parameter between these players.