Brief Review

A Review of the Validity and Reliability of

Accelerometer-Based Metrics From Upper
Back–Mounted GNSS Player Tracking Systems for
Athlete Training Load Monitoring
Laura Dawson,1,2 Marco Beato,1 Gavin Devereux,1 and Stuart A. McErlain-Naylor1,3
1
School of Allied Health Sciences, University of Suffolk, Ipswich, United Kingdom; 2Faculty of Sport, Technology and Health Sciences,
St Mary’s University, Twickenham, United Kingdom; and 3School of Sport, Exercise and Health Sciences, Loughborough University,
Loughborough, United Kingdom

Abstract
Dawson, L, Beato, M, Devereux, G, and McErlain-Naylor, SA. A review of the validity and reliability of accelerometer-based metrics
from upper back–mounted GNSS player tracking systems for athlete training load monitoring. J Strength Cond Res 38(8):
e460–e475, 2024—Athlete load monitoring using upper back–mounted global navigation satellite system (GNSS) player tracking is
common within many team sports. However, accelerometer-based load monitoring may provide information that cannot be
achieved with GNSS alone. This review focuses on the accelerometer-based metrics quantifying the accumulation of accelerations
as an estimation of athlete training load, appraising the validity and reliability of accelerometer use in upper back–mounted GNSS
player tracking systems, the accelerometer-based metrics, and their potential for application within athlete monitoring. Reliability of
GNSS-housed accelerometers and accelerometer-based metrics are dependent on the equipment model, signal processing
methods, and the activity being monitored. Furthermore, GNSS unit placement on the upper back may be suboptimal for
accelerometer-based estimation of mechanical load. Because there are currently no feasible gold standard comparisons for field-
based whole-body biomechanical load, the validity of accelerometer-based load metrics has largely been considered in relation to
other measures of training load and exercise intensity. In terms of convergent validity, accelerometer-based metrics (e.g., Play-
erLoad, Dynamic Stress Load, Body Load) have correlated, albeit with varying magnitudes and certainty, with measures of internal
physiological load, exercise intensity, total distance, collisions and impacts, fatigue, and injury risk and incidence. Currently,
comparisons of these metrics should not be made between athletes because of mass or technique differences or between
manufacturers because of processing variations. Notable areas for further study include the associations between accelerometer-
based metrics and other parts of biomechanical load-adaptation pathways of interest, such as internal biomechanical loads or
methods of manipulating these metrics through effective training design.
Key Words: PlayerLoad, Dynamic Stress Load, Body Load, acceleration, biomechanical load

Introduction internal load measures, and training outcomes (100,114,144,201).

These frameworks can be informed by research in physiology,
Athletic training can have various intended and unintended effects
biomechanics, and other areas (7,100,107,108). Determining ap-
on the human system. The investigation, prescription, and moni-
propriate measures of training load can provide application in the
toring of these cause-and-effect relationships require the training
and the outcome to be quantified. “Training load” is considered prescription and monitoring of training sessions within an overall
a multidimensional construct consisting of 2 causally related sub- framework of physical training (72,74,102) and inform subsequent
dimensions: “external training load” and “internal training load” session design based on differences between prescribed and ach-
(98–100), each of which can be subdivided into physiological or ieved training loads (44).
biomechanical load. For example, external physiological load The past 2 decades have seen rapidly increasing use of wearable
could include kinetic energy (i.e., distance covered), whereas in- technologies to monitor athlete training loads (33,88). However,
ternal physiological load could be oxygen uptake or cardiovascular there is a need for practitioners to first refine their requirements
demand (189). External biomechanical load may include ground for a training load monitoring system within any causal frame-
reaction forces (GRF) or center of mass (CoM) accelerations, work and then evaluate available devices against those require-
whereas internal biomechanical load may refer to joint contact ments (8,37). One of the most well-known and commonly used
forces or muscle-tendon forces (189). Identification of an appro- areas of technology for this purpose is global positioning system
priate measure or metric of training load ideally requires a frame- (GPS) or global navigation satellite system (GNSS) technology
work of causal relationships between external load measures, (8,20–22,49). Global positioning system is 1 subsection of GNSS,
with modern GNSS accessing a greater number of satellites and
Address correspondence to Laura Dawson, [email protected]. therefore potentially providing greater precision and reliability
Journal of Strength and Conditioning Research 00(00)/e1–e16 (101,123). Both GPS and GNSS are considered collectively in this
ª 2024 National Strength and Conditioning Association article as GNSS.

e1

Copyright © 2024 National Strength and Conditioning Association. Unauthorized reproduction of this article is prohibited.
 A Review of the Validity and Reliability of Accelerometers (2024) 00:00

A GNSS player tracking unit can accurately report total and accelerometer-based metrics estimating external biomechanical
high-speed distance covered, and peak and average speeds (21) load from the accumulation of instantaneous accelerations. Spe-
but underestimates low-velocity high-impact events with minimal cifically, this review aims in assessing the validity, reliability, and
horizontal displacement, such as tackling, jumping, landing, potential applications in monitoring athlete load, training pre-
blocking, and other collisions (5,28). Frequent changes of di- scription, and injury management as a general concept, through
rection can also affect the validity and reliability of some GNSS accumulating acceleration measurements from GNSS-housed
metrics, such as total distance (140,155) and instantaneous ve- accelerometers. This process involves applying manufacturer-
locity (94). Applications of GNSS are further limited in indoor specific algorithms to the accelerometer signal, reported as arbi-
sports, although local positioning systems can be used (147,171) trary units. Examples include Player Load (Catapult, Melbourne,
and when activity is performed with relatively little translational Australia), Dynamic Stress Load (STATSports, Newry, Northren
movement (192). Consequently, an inertial measurement unit Ireland) and Body Load (GPSports, Canberra, Australia).
(IMU) comprising accelerometer, gyroscope, and magnetometer
is often incorporated into the GNSS unit (9,58). The accelerom-
eter provides the capability to monitor external biomechanical Accelerometers
load through instantaneous accelerations experienced by the
athlete (123,193). As the trunk is the heaviest body segment and Accelerometers measure linear acceleration and are typically
positioned proximally on the body, these accelerations have been uniaxial or triaxial, measuring accelerations in 1 or 3 (e.g.,
proposed to represent whole-body CoM accelerations and anterior-posterior, mediolateral, and vertical) axes, respectively
therefore relate to the GRF or impact magnitudes experienced. (203). Triaxial microelectromechanical system accelerometers
However, it may not be possible to accurately predict whole-body are the most common in sport, measuring acceleration magni-
CoM accelerations and hence GRF from accelerations of a single tudes in each axis (9,82). These devices can collect data at high
or small number of segments (61,137,138,190,203). Such exter- frequencies, are sensitive to rapid changes in acceleration (185),
nal forces describe the biomechanical loading experienced by the and can be used in indoor venues because of the lack of a re-
musculoskeletal system and have been commonly investigated in quirement for satellite connection.
association with running performance (26,136,150) and injuries Accelerometers are typically used to measure accelerations of
(188,209). However, relationships between GRF and musculo- a body segment (e.g., torso, shank) or an object (e.g., racket, stick)
skeletal injuries are often unclear (188,209), and these external (40). Investigations into optimal accelerometer placement have
forces may not reflect internal biomechanical loads (126,210). identified CoM (70,110,115,122) or ankle attachments
From a training perspective, the monitoring of instantaneous (53,90,125) to be most effective for estimating GRF and thus the
accelerations is particularly useful for assessing external load in external biomechanical load experienced by the athlete
collision sports, such as rugby and some codes of football, given (70,80,92,122). This is perhaps unsurprising due to the funda-
the frequency of impacts (93,96,106). Accelerometer-based mental relationship between GRF and CoM acceleration
metrics include peak acceleration, number of accelerations (69,111). However, accelerometers embedded in GNSS units are
above a specified threshold, resultant vector quantities, and cu- mounted on the upper back for optimal GNSS signal acquisition
mulative acceleration metrics that estimate mechanical load (e.g., (18,123) and athlete comfort. Practitioners should consider that
PlayerLoad, Dynamic Stress Load, Body Load) (30,153,189). upper back–mounted accelerometer signals can be affected by
Although coaches, sport scientists, and medical teams are using factors, such as incidental harness or attachment movement
accelerometer-based metrics to monitor components of perfor- (61,132,206); therefore, sensor fixation should be as tight as is
mance, training load, fatigue, and injury risk (3,43,49,120,200), physically comfortable to minimize additional displacement
it is less apparent which metrics have received adequate assess- (32,95,118,186). In addition, trunk positioning (24,68) and gait
ment of validity and reliability (30,97,121,139). To optimize cycle features, including cadence, step length, and gait regularity
training for a specific outcome within a causal framework, any and symmetry (134,159), can affect the magnitude and timing of
applied measure of training load should reflect the mediating the postimpact elastic wave (propagation of impact-related en-
mechanisms of the primary outcome (45,100,114). Therefore, the ergy [i.e., measured accelerations] through body tissues and de-
purpose of validating training load metrics is to assess the extent creasing of amplitude with distance from the source). Finally,
to which this is true for any given metric (100). If a training load muscle damping by which sites further from impact are also less
measure cannot be connected causally to a plausible mediator of affected by the amplifying effect of greater GRF (35,129,194)
the desired outcome, then it is likely of little practical use for Therefore, the correlations between accelerations or
purposes of training optimization (100). Because of the numerous accelerometer-based metrics at different body positions may not
performance or injury outcomes of interest, each with their own be particularly high (80). It should also be considered that ac-
set of mediators, there cannot be a single best metric that reflects celeration signals comprise various frequencies, not all of which
all causal pathways from athletic training to its outcomes. Ulti- relate directly to the postimpact elastic wave (117,129,174).
mately, the appropriateness of a metric may depend on what
aspects of a causal framework it is intended to describe or predict,
Accelerometer-Based Load Metrics
its relationships with other aspects of the causal framework, the
acceptability of its limitations, and whether the assumptions on Sport and exercise science researchers and practitioners have
which it is based are reasonable or testable (100). Given that typically sought to convert continuous signals (e.g., an accelera-
researchers and practitioners are using these metrics, it is im- tion time-series signal) into discrete metrics such as mean or peak
portant to know what they can and cannot do in a valid and values, and durations (100). This often sees intensity and duration
reliable manner as part of a causal framework. combined to operationalize a single cumulative training load
Therefore, the aim of this review is to appraise the literature metric (39,100). However, practitioners often prefer automated
regarding the validity and reliability of accelerometer use in upper metrics from which meaning can be relatively easily inferred,
back–mounted GNSS player tracking systems, with a focus on rather than continuous data from an entire session that requires

e2

Copyright © 2024 National Strength and Conditioning Association. Unauthorized reproduction of this article is prohibited.
 A Review of the Validity and Reliability of Accelerometers (2024) 00:00 | www.nsca.com

complex additional processing and lacks direct application. 0.15-m single-leg drop landings may not have been sufficient to
Metrics would ideally represent an underlying (e.g., mechanical) exceed the abovementioned thresholds in some individuals. This
phenomena and relate to causal mechanisms of injury risk, tissue is an area that speaks to the requirement for further research to
remodeling, or training adaptations that can be tested as part of better understand what is measured and included in the approx-
a causal pathway through transparent calculation and processing imation or accumulation of accelerations expressed in these ac-
methods. Once an appropriate measure of training load is iden- celerometer load metrics. Practitioners would be better able to
tified, it is necessary to determine the relevant dimensions make informed decisions with a better understanding of the
(i.e., metrics) for its quantification and application (45,100). quantification of these arbitrary unit metrics.
Table 1 summarizes what is known of the equation and calcula-
tion process for 3 of the most common accelerometer-based
metrics: PlayerLoad, Body Load, and Dynamic Stress Load. Validity and Reliability
Since the introduction of PlayerLoad, the definition and
equation used have become less clear. One study found Catapult’s The manufacturers and metrics will be considered individually
automatically derived PlayerLoad values to be 15.1 6 1.2% because of differences in the algorithms, unknown variations in
lower than those calculated manually using the stated equation signal processing before metric calculation, or differences caused
(139). Bredt et al. (30) identified 4 differing definitions and 4 by the application of the same algorithm to different manu-
differing equations for PlayerLoad between publications facturers’ accelerometers (139). The validity is therefore context
(2,27,34,36,154,165,198). Hollville et al. (91) claimed to have specific, in that any metric can be assessed for a specific purpose
determined the manufacturer processing methods during a pilot but not extrapolated to all possible applications of that metric
analysis as low-pass filtering with a zero-phase-shifting third- (64,170). In addition, data validity will depend on the technical
order Butterworth filter at 1 Hz, applying a scaling factor of 4. specifications of the sensor, the accuracy of sensor calibration,
Despite these inconsistencies, PlayerLoad has been the most sensor placement and fixation, and signal processing methods,
commonly researched metric (81,189). among other factors (95).
Variation also exists in the equations reported for GPSports’ Within a causal framework, validation requires that the metric
Body Load metric, as demonstrated in Table 1. As a result, it is not is a valid reflection of the proposed causal mechanism linked to
always clear which version of the metric has been used within the the outcome (i.e., construct validity) and that the proposed causal
literature. GPSports are now under the Catapult umbrella, and mechanism is indeed causally related to the outcome of interest
Body Load is less frequently used, but devices and systems remain (100). Practitioners should additionally be aware of the reliability
in circulation. Research using Body Load can still inform our of the measurement system consisting of device, processing, and
understanding of accelerometer-based metrics and their calculation (169). Differences between individuals, sessions, or
capabilities/limitations within causal frameworks of interest. drills can only meaningfully be interpreted within the context of
The 3 metrics differ in their apparent underlying assumptions intraunit or interunit and intraindividual or /interindividual re-
or their operationalization of the training load concept. Body liability, which should help inform best practice for data collec-
Load includes individual acceleration values, similar in some tion and processing methods in both research and applied settings
ways to the calculation of GRF impulse; Dynamic Stress Load (1,10). As a result, it is important to review the validity of raw
includes peak accelerations, similar in some ways to the calcula- acceleration measures; the reliability of raw acceleration meas-
tion of peak GRF; and PlayerLoad includes the change in accel- ures; the reliability of any metrics calculated from the measured
eration, similar in some ways to the calculation of GRF loading accelerations; how these metrics relate to other parts of any po-
rate. In reality, each metric is more complex than that with mul- tential performance or injury-related causal framework; what
tiple components, scaling factors, and nonlinear components. such metrics can or cannot do as part of any such applied
Nonetheless, it remains important to consider how the underlying framework; and what additional research is necessary to inform
equations may reflect any hypothetical relationships between the design and application of these frameworks.
“load” and subsequent outcomes of interest. As recently stated by The intraclass correlation coefficient (ICC) assesses the agree-
Impellizzeri et al. (100), it appears that the “burden of proof” has ment between data (e.g., interunit, interday, relationship to a gold
been reversed; non-theory-driven metrics of training load are standard measure) (112). The coefficient of variation (CV) is the
being commercially presented, requiring researchers and practi- ratio of the standard deviation to the mean (172). Throughout
tioners to investigate their relevance within any causal frame- this review, ICC and other magnitudes of correlation will be
works of interest. There is often no explicitly stated theoretical interpreted as 0.0 # negligible , 0.3 # low , 0.5 # moderate ,
framework underlying the equations used or linking them to the 0.7 # high , 0.9 # very high # 1.0 (89,135). Coefficient of
mechanisms they are proposed to reflect. variation will be interpreted as good # 5% , moderate # 10% ,
The metrics also differ in their use, or not, of a particular poor (168). However, the importance of context-specific evalu-
threshold. For example, Dynamic Stress Load includes all peak ation such as absolute differences and their practical significance
resultant accelerations $ 2 g, whereas Body Load includes all time for the intended application within a causal framework should
points where resultant acceleration $ 1.25 g. This may be an not be underestimated (87).
attempt to avoid the inclusion of random noise or very low
impacts within the cumulative calculations. The specificity of
Validity of Upper Back–Mounted Global Navigation Satellite
these thresholds to the between-scapulae measurement site should
System–Housed Accelerometer Data
be considered when interpreting studies that have applied the
same processes at alternative body sites. For example, single-leg Any upper back–mounted accelerometer should be evaluated
drop landings from a height of 0.45 m have resulted in peak against either true acceleration of the unit or an estimated accel-
surface-measured accelerations on average of 68 g at the meta- eration of the underlying structure it is intended to represent as
tarsophalangeal joint but 4 g at the C6 vertebra (129). Peak part of the causal framework (e.g., trunk segment or whole-body
accelerations at the C6 vertebra level of 2.6 6 0.8 g following CoM). Typically, upper back–mounted accelerometers have been

e3

Copyright © 2024 National Strength and Conditioning Association. Unauthorized reproduction of this article is prohibited.
Copyright © 2024 National Strength and Conditioning Association. Unauthorized reproduction of this article is prohibited.

A Review of the Validity and Reliability of Accelerometers (2024) 00:00

Table 1
Manufacturer-designed metrics derived from the triaxial accelerometers embedded in upper back–mounted global navigation satellite system player tracking technology. (g,
acceleration due to gravity, 9.81 m·s22).
Metric (manufacturer) Equation Details of calculation
PlayerLoad (Catapult Sports) Described as the square root of the sum of changes between adjacent time points in
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi acceleration in all 3 individual axes, divided by a scaling factor of 100 to give
ðaxn 2 axn 2 1 Þ2 1 ðayn 2 ayn 2 1 Þ2 1 ðazn 2 azn 2 1 Þ2 a summation in arbitrary units (30,86,138,192)
PL ¼ +
100
where ay,ax,az 5 acceleration in the x,y,z axes, and n 5 the
time point (71,138)
Body Load (GPSports) Normalized Acceleration Vector ðNAVÞ Some early studies report using the aforementioned PlayerLoad equation (2,87). In
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi others, a multiplication by subject body mass and a loading factor was used instead of
¼ ax 2 1 ay 2 1 az 2 2 g the division using a denominator of 100 (52,122). The loading factor ensured
a nonlinear relationship between trunk-measured accelerations and Body Load, as
this reflected the magnitude of resultant accelerations, e.g., 5.0 g # 6.0 g [light
impact] 5 1; 6.0 g # 6.5 g 5 2; 6.5 # 7.0 g [moderate impact] 5 3; 7.0 # 8.0 g
e4

[heavy impact] 5 4; 8.0 # 10.0 g 5 5; .10.0 g [severe impact] 5 6

The manufacturers have since proposed “New Body Load,” which subtracts g from
  the resultant acceleration before calculation, includes a linear and nonlinear
NAV 1 NAV 3 4frequency component, and is scaled by both the accelerometer sampling frequency and an
Body Load ¼ +
Exercise Factor unknown manufacturer-determined “exercise factor” (83,146,191). This is summed
for all time points with a resultant acceleration $1.25 g and reported in arbitrary units
where ax, ay, az 5 acceleration in the x,y,z axes
Dynamic Stress Load (STATSports) Dynamic Stress Load calculates the accumulation of acceleration as a measure of
n
athlete load in arbitrary units, calculated from “the weighted sum of all impacts above
DSL ¼ + peak accki 3 SF or equal to 2 g” (20,22,193). These “impacts” are classified as “maximum
i¼1 accelerometer magnitude values above 2 g in a 0.1 s period” to isolate the peak
acceleration of each ground contact above the stated intensity threshold. The values
where peak acc is the maximum resultant acceleration $2 g are weighted using a nonlinear function, based on an assumed nonlinear relationship
in a 0.1 s period. k is a weighting factor and SF is a scaling between impact magnitude and the consequences of that impact within the body; for
factor—both are proprietary factors of the manufacturer (20) example, this assumes that a peak acceleration of 4 g is associated with more than
twice the consequences of a 2 g peak acceleration. Instantaneous values are summed
to provide Dynamic Stress Load (DSL) over defined time periods (78,83)
 A Review of the Validity and Reliability of Accelerometers (2024) 00:00 | www.nsca.com

Table 2
Studies reporting validity of global navigation satellite system–housed accelerometers using raw and filtered raw data.*†
Article Device Criterion Activity Filter cutoff Error Correlation
Edwards et al. GPSports SPI HPU Full-body marker set—3D All Manufacturer proprietary WA: CV 5 12.68% ICC 5 0.22
(61) (100 Hz) motion capture system (300 Hz) filtering (poor) (negligible)
P: CV 5 7.62% ICC 5 0.61
(moderate) (moderate)
Slow running (3.3 m·s21) WA: CV 5 6.19% ICC 5 0.89
(moderate) (high)
P: CV 5 6.06% ICC 5 0.86
(moderate) (high)
Medium running (5.0 m·s21) WA: CV 5 5.12% ICC 5 0.94
(moderate) (very high)
P: CV 5 6.10% ICC 5 0.85
(moderate) (high)
Fast running (6.7 m·s21) WA: CV 5 16.20% ICC 5 0.72
(moderate) (high)
P: CV 5 13.12% ICC 5 0.78
(poor) (high)
Kelly et al. GPSports SPI-ProX II ADXL345 triaxial sensor (100 Hz) Static (flat surface with no more Raw %Diff 5 27.5 to
(109) (100 Hz) than 2˚ axis deviation) 30.5
Dynamic (fixed oscillation point of Raw %Diff 5 32.0 to
a vibration platform) 35.0
Nedergaard Catapult MinimaxX Ground reaction force–derived Straight line run Raw r 5 0.26
et al. (137) S4 (100 Hz) centre of mass acceleration (negligible)
45˚ CoD Raw r 5 0.42 (low)
90˚ CoD Raw r 5 0.55
(moderate)
Roell et al. Catapult OptimEye Three markers on device—3D Sport-specific movements (series Complementary (accel. and
(162) S5 (100 Hz) motion capture system (200 Hz) of drills involving acceleration gyroscope) filter
and deceleration, 60˚, 80˚, 90˚, 100 Hz Mean: CV 5 7.34% r 5 0.99 (very
and 360˚ CoD, and jumps) (moderate) high)
Peak: CV 5 10.05% r 5 0.97 (very
(poor) high)
10 Hz Mean: CV 5 7.46% r 5 0.98 (very
(moderate) high)
Peak: CV 5 10.63% r 5 0.97 (very
(poor) high)
5 Hz Mean: CV 5 r 5 0.96 (very
10.78% (poor) high)
Peak: CV 5 15.98% r 5 0.95 (very
(poor) high)
Kalman Filter
100 Hz Mean: CV 5 5.99% r 5 0.99 (very
(moderate) high)
Peak: CV 5 12.10% r 5 0.95 (very
(poor) high)
10 Hz Mean: CV 5 8.90% r 5 0.98 (very
(moderate) high)
Peak: CV 5 12.74% r 5 0.94 (very
(poor) high)
5 Hz Mean: CV 5 r 5 0.90 (very
16.48% (poor) high)
Peak: CV 5 22.56% r 5 0.85 (high)
(poor)
Wundersitz Catapult MinimaxX Single marker on unit—3D Walk (1.5 m·s21) 10 Hz CV 5 6.5%
et al. (204) S4 (100 Hz) motion capture system (200 Hz) (moderate)
Jog (3.3 m·s21) 10 Hz CV 5 7.5%
(moderate)
Run (5.0 m·s21 female, 5.9 10 Hz CV 5 9.3%
m·s21 male) (moderate)
Walk-jog-run Raw CV 5 15.8% (poor)
20 Hz CV 5 13.8% (poor)
15 Hz CV 5 11.5% (poor)
10 Hz CV 5 8.9%
(moderate)
8 Hz

e5

Copyright © 2024 National Strength and Conditioning Association. Unauthorized reproduction of this article is prohibited.
 A Review of the Validity and Reliability of Accelerometers (2024) 00:00

Table 2
Studies reporting validity of global navigation satellite system–housed accelerometers using raw and filtered raw data.*† (Continued)
Article Device Criterion Activity Filter cutoff Error Correlation
CV 5 8.2%
(moderate)
6 Hz CV 5 8.2%
(moderate)
Wundersitz Catapult MinimaxX Single marker on unit—3D Physical collisions Raw CV 5 15.1% (poor)
et al. (205) S4 (100 Hz) motion capture system (500 Hz) 30 Hz CV 5 12.0% (poor)
25 Hz CV 5 10.8% (poor)
20 Hz CV 5 9.6%
(moderate)
15 Hz CV 5 9.5%
(moderate)
10 Hz CV 5 14.5% (poor)
8 Hz CV 5 19.3% (poor)
*CoD 5 change of direction; WA 5 weight acceptance phase; P 5 propulsion phase; %Diff 5 percentage difference.
†Magnitude of correlations interpreted as 0.0 # negligible , 0.30; 0.3 # low , 0.50; 0.50 # moderate , 0.70; 0.70 # high , 0.90; very high $ 0.90 (93,139). Coefficient of variation (CV) interpreted as
good # 5% , moderate # 10% , poor (173).

evaluated against 3D motion capture (61,204,205). Table 2 purpose is not to track segmental kinematics but to monitor the
summarizes studies evaluating the validity of upper back- postimpact elastic wave accelerations, filtering decisions should
–mounted accelerometers based on correlations between raw or perhaps be made in relation to the frequency content of the signal
filtered accelerometer data and 3D motion capture. Roell et al. component of interest (84,103,129,174).
(162) demonstrated that applying a complementary filter (com-
bining accelerometer and gyroscope data) and resampling to 5 Hz
decreased resultant acceleration root mean square error by 15% Reliability of Global Navigation Satellite System–Housed
(0.52–0.44 m·s22) for mean and 58% (3.30–1.40 m·s22) for peak Upper Back–Mounted Accelerometer Data
values in sport-specific movements. In summary, raw signal data
Comparatively, studies investigating the reliability of GNSS-
from upper back–mounted GNSS units are generally less valid
housed upper back–mounted accelerometer raw data are rela-
than filtered signal data for estimating trunk accelerations when
tively scarce, with most studies preferring to evaluate manufac-
assessed against 3D motion capture (Table 2) (89,135).
turer accelerometer-based metrics. This could be attributed to the
In addition, studies have assessed the capabilities of GNSS-
relative ease of obtaining and interpreting manufacturer-
embedded upper back–mounted accelerometers using correla-
processed data as opposed to raw accelerometer data from the
tions to GRF (Table 3), evaluating the relationship between ac-
GNSS units. Nonetheless, Table 4 displays studies that have
celerometer data and whole-body external biomechanical load
reported the reliability of data from GNSS-housed accel-
(i.e., GRF parameters). It is noteworthy that the trunk represents
erometers. Overall, the reliability of GNSS-housed accelerometer
the largest proportion of body mass (;50%), perhaps explaining
data and metrics is largely high to very high, with predominantly
why the segmental acceleration of the trunk often best represents
good CV. However, some findings demonstrated only moderate
the acceleration of whole-body CoM (137). Collectively, these
correlations and moderate to poor CV. Therefore, practitioners
studies demonstrate that although linear relationships between
should always consider the potential implications of the type of
upper back–mounted accelerations and whole-body external
activity (e.g., static, dynamic, field-based) and surface (e.g.,
biomechanical load may be present in team sports after appro-
treadmill, track) on accelerometer data (27,79,80,109). If the
priate processing (47), they are weak in some cases and so
context-specific validity and reliability of acceleration data are
assumptions of simple linear relationships should be made with
deemed acceptable, these can then be used to calculate training
caution (137). If a measure of upper back–mounted acceleration
load metrics, which should themselves be assessed for reliability.
is valid for its intended purpose, subsequent inferences depend on
Table 5 displays studies reporting the reliability of
the reliability of those measurements. Any training load algo-
accelerometer-based metrics, demonstrated by controlled shaker-
rithms will ideally be applied to valid and reliable input data.
table, standardized treadmill running, multidirectional soccer-
Overall, low-pass filtering at 8 Hz or 10 Hz for walking and
specific tasks, and standardized soccer games. This review shows
jogging (CV: 6.5–7.5%, moderate) (204), 15 Hz or 20 Hz for
Catapult’s PlayerLoad metric has been the most extensively
physical collisions (CV: 9.5% or 9.6%, moderate) (205), and
assessed accelerometer-based metric for reliability. Overall find-
20 Hz for continuous hopping, drop landing, and rebound jumps
ings suggest that PlayerLoad and some of its derived metrics can
(r 5 0.709 to 0.825, high) (176) have all been shown to improve
be reliable provided that the same device is used for repeated tests
the relationship between upper back–mounted accelerometers
(15,16,139).
and 3D motion capture (204,205) or GRF (176). The general
indication is that upper back–mounted accelerometers embedded
in GNSS units can provide valid estimations of trunk or CoM
Convergent Validity of Accelerometer-Based Metrics
accelerations when a complementary filter is applied or when the
signal is filtered at 20 Hz for high-impact activities and 10 Hz for It can be difficult to assess the validity of PlayerLoad, Dynamic
lower-impact activities. However, filtering decisions are again Stress Load, Body Load. or similar accelerometer-based metrics
dependent on the purpose of data collection and desired rela- against internal biomechanical load during sporting activities,
tionships with other factors within an overall framework. If the given that a true gold standard method would require invasive

e6

Copyright © 2024 National Strength and Conditioning Association. Unauthorized reproduction of this article is prohibited.
 A Review of the Validity and Reliability of Accelerometers (2024) 00:00 | www.nsca.com

measures of continuous loading of internal structures. However, may be more valid for some specific purposes when the larger
existing literature has attempted to validate some metrics using vertical acceleration component cannot mask smaller changes in
comparisons to variations of the same metric (to identify key other axes, such as horizontal collision accelerations masked by
features of the algorithm in a sensitivity analysis or determine the running foot strikes (50,55,132), or when the purely rotational
most effective version for a specific purpose) or assessed agree- components are disregarded (91). This may need to be considered
ment relative to similar metrics. For example, Wylde et al. (206) in contexts where horizontal accelerations are of primary interest;
compared PlayerLoad data from VXSport Log accelerometers however, it does need further research to be fully understood.
with Catapult Optimeye S5 accelerometers (both 100 Hz), finding Further to these studies, PlayerLoad, Body Load, and Dynamic
high to very high correlations between the devices for vertical, Stress Load have all been assessed for correlation with various
anterior-posterior, and mediolateral PlayerLoad (ICC: biomechanical and physiological external and internal load
0.970–0.956, 0.828; high to very high) (100). measures. Consequently, this review considers the convergent
Two related studies assessing tackling impacts in rugby league validity of these metrics, referring to how closely one metric is
highlighted lower sensitivity to changes in mediolateral and associated with other measures of the same construct (e.g., other
anteroposterior axis accelerations because of high vertical accel- internal and external training load measures or parameters that
erations (50,132). Although Hollville et al. (91) recognized the one would expect to be associated with a valid measure of bio-
potential for limitations in the PlayerLoad algorithm given that it mechanical training load). This can inform decisions regarding
may be influenced by rotational changes of the reference frame the validity of that metric as an operationalization of the concept
potentially leading to erroneous calculations. These Body Load of biomechanical training load, although interpreting the metrics
and PlayerLoad studies suggest that accelerometer-based metrics as true mechanical constructs should be cautioned against.

Table 3
Studies reporting correlations between raw and filtered global navigation satellite system–housed accelerometer data and ground
reaction force parameters using raw and filtered data.*†
Article Device Criterion Activity Filter cutoff CV (%) Correlation
Edwards et al. (61) GPSports SPI HPU (100 Vertical ground reaction force All Manufacturer-specific WA: 12.68 (poor) ICC 5 20.11
Hz) (1200 Hz) filter (negligible)
P: 7.62 ICC 5 20.32 (low)
(moderate)
Slow running (3.3 WA: 6.19 ICC 5 0.47 (low)
m·s21) (moderate)
P: 6.06 ICC 5 20.36 (low)
(moderate)
Medium running (5.0 WA: 5.12 ICC 5 0.50
m·s21) (moderate) (moderate)
P: 6.10 ICC 5 20.15
(moderate) (negligible)
Fast running (6.7 WA: 16.20 (poor) ICC 5 0.31 (low)
m·s21) P: 13.12 (poor) ICC 5 20.02
(negligible)
Simons and Bradshaw Catapult MinimaxX S4 Vertical ground reaction force Continuous hopping 50 Hz r 5 0.818 (high)
(176) (100 Hz) (500 Hz) 20 Hz r 5 0.825 (high)
15 Hz r 5 0.825 (high)
8 Hz r 5 0.860 (high)
Drop landing 50 Hz r 5 0.699
(moderate)
20 Hz r 5 0.727 (high)
15 Hz r 5 0.720 (high)
8 Hz r 5 0.746 (high)
Rebound jump 50 Hz r 5 0.734 (high)
20 Hz r 5 0.734 (high)
15 Hz r 5 0.706 (high)
8 Hz r 5 0.613
(moderate)
Wundersitz et al. (203) GPSports SPI Pro (100 Hz) Resultant ground reaction force Straight line run Raw 16.4 (poor) r 5 0.31 (low)
(100 Hz) 10 Hz 11.7 (poor) r 5 0.76 (high)
45 CoD Raw 16.6 (poor) r 5 0.00 (negligible)
10 Hz 14.5 (poor) r 5 0.67 (moderate)
90 CoD Raw 18.1 (poor) r 5 0.48 (low)
10 Hz 17.2 (poor) r 5 0.47 (low)
180 CoD Raw 23.0 (poor) r 5 0.30 (low)
10 Hz 23.9 (poor) r 5 0.23 (negligible)
*CoD 5 change of direction; WA 5 weight acceptance phase; P 5 propulsion phase.
†Magnitude of correlations interpreted as 0.0 # negligible , 0.30; 0.3 # low , 0.50; 0.50 # moderate , 0.70; 0.70 # high , 0.90; very high $ 0.90 (93,139). Coefficient of variation (CV) interpreted as
good # 5% , moderate # 10% , poor (173).

e7

Copyright © 2024 National Strength and Conditioning Association. Unauthorized reproduction of this article is prohibited.
 A Review of the Validity and Reliability of Accelerometers (2024) 00:00

Table 4
Studies reporting reliability of global navigation satellite system–housed accelerometer data.*
Acceleration
Article Device Activity measure CV (%) Correlation
Gomez-Carmona et al. RealTrack WIMU PRO (1000 Between-device Resultant 0.38 (good)
(81) Hz) Static 0.78 (good)
Static with active intervals 0.59 (good)
Dynamic—hydraulic shaker 10 Hz 0.66 (good)
Dynamic—hydraulic shaker 30 Hz 2.05 (good)
Treadmill incremental running 2.46 (good)
SAFT90 soccer simulation drill
Within-Device Resultant 0.23 (good)
Static 0.40 (good)
Static with active intervals 0.55 (good)
Dynamic—hydraulic shaker 10 Hz 0.12 (good)
Dynamic—hydraulic shaker 30 Hz
Kelly et al. (109) GPSports SPI-ProX II (100 Mechanical device capable of reproducing Raw peak 1.87–2.21
Hz) highly consistent impacts acceleration (good)
Nicolella et al. (139) Catapult OptimEye S5 (100 Anterior-posterior Peak
Hz) Shaker table oscillating at 3.0 g ICC 5 1.00 (0.99–1.00) (very high)
Shaker table oscillating at 1.0 g ICC 5 1.00 (0.99–1.00) (very high)
Shaker table oscillating at 0.5 g ICC 5 1.00 (0.99–1.00) (very high)
Shaker table oscillating at 0.1 g ICC 5 1.00 (0.99–1.00) (very high)
Mediolateral Peak
Shaker table oscillating at 3.0 g ICC 5 0.77 (0.62–0.89) (moderate to
high)
Shaker table oscillating at 1.0 g ICC 5 0.96 (0.92–0.98) (very high)
Shaker table oscillating at 0.5 g ICC 5 1.00 (0.99–1.00) (very high)
Shaker table oscillating at 0.1 g ICC 5 1.00 (0.99–1.00) (very high)
Vertical Peak
Shaker table oscillating at 3.0 g ICC 5 1.00 (1.00–1.00) (very high)
Shaker table oscillating at 1.0 g ICC 5 1.00 (1.00–1.00) (very high)
Shaker table oscillating at 0.5 g ICC 5 1.00 (1.00–1.00) (very high)
Shaker table oscillating at 0.1 g ICC 5 1.00 (1.00–1.00) (very high)
*Magnitude of correlations interpreted as 0.0 # negligible , 0.30; 0.3 # low , 0.50; 0.50 # moderate , 0.70; 0.70 # high , 0.90; very high $ 0.90 (93,139). Coefficient of variation (CV) interpreted as
good # 5% , moderate # 10% , poor (173).

Relationships With External Load Measures. Relationships be- Australian football (11,17), rugby league (73,97), rugby union
tween accelerometer-based metrics and exercise demand has been (160), soccer (34), and hockey (148). This is likely at least partly
thoroughly researched. The correlations reported in this review a result of an increased number of foot strikes when covering
for the metrics and algorithms demonstrate largely moderate to greater distances. Conversely, there is a very limited body of lit-
very high correlations with known external load measures sug- erature assessing the relationship between derivatives of Dynamic
gesting the potential for acceleration accumulation metrics to be Stress Load and distance; 1 study reports no correlation
applied effectively for specific purposes within athlete load (r 5 20.01) with distance covered per minute in a study of 28
monitoring. There is evidence of Dynamic Stress Load being ca- male futsal players (158). Current findings imply that factors
pable of distinguishing low, medium, and high training session other than running distance are likely determinants of
intensities (67), increased load in competition compared with accelerometer-based load metrics in sports such as futsal, con-
training (181), and a small difference between first team players sisting of frequent high-intensity accelerations, decelerations, and
and under-18 and under-23 age group players (157). Meanwhile, sport-specific actions that do not require running for long dis-
PlayerLoad has been shown to increase with treadmill running tances (13,116).
speed (16), the number and magnitude of rugby league collisions
(71), and distance covered in soccer training (34). Together, this Running Accelerations and Decelerations. Accelerations and
research suggests the potential for accelerometer-based metrics to decelerations in running speed, beginning at 1 m·s22 and ex-
distinguish between intensity levels in varied sports and activities, ceeding 2 m·s22 for at least half a second, are recognized con-
indicative of a certain level of convergent validity. However, tributors to overall mechanical load (85,86). Accelerations and
researchers and practitioners should always consider their choice decelerations have respectively contributed 7–10% and 5–7% of
of measurement device or metric in relation to their specific a total “player load” (a downscaled square of resultant acceler-
objective. ation) across all playing positions during 45 elite competitive
soccer matches (54). It has been previously found that GPSports
Total Distance. Distance covered has been adopted as a perfor- raw accelerometer data consistently overestimated average high-
mance and load monitoring parameter, used to assess playing speed run acceleration from change in speed over 0–10 m and
standard (31,105), aerobic capacity (151,164), fatigue 10–20 m of a 20-m distance. This was attributed to the need to
(6,133,211), and training loads (152,211). Multiple studies have incorporate body mass and compensate for gravity to improve the
assessed the relationship between PlayerLoad and distance cov- accuracy and practicality of accelerometer-derived algorithms
ered, finding high to very high correlations (r 5 0.70 to 0.98) in (5). However, Beato and Drust (20) found that maximal

e8

Copyright © 2024 National Strength and Conditioning Association. Unauthorized reproduction of this article is prohibited.
 A Review of the Validity and Reliability of Accelerometers (2024) 00:00 | www.nsca.com

Table 5
Studies reporting reliability of accelerometer-based load metrics.*†
Article Device Activity Metric CV (%) Correlation
Barreira et al. (15) STATSports Viper (100 Hz) Jogging PL 14.5 (poor) r 5 0.863 (high)
PL·min21 18.2 (poor) r 5 0.903 (very high)
Side cut PL 15.2 (poor) r 5 0.892 (high)
PL·min21 17.8 (poor) r 5 0.921 (very high)
Stride PL 24.5 (poor) r 5 0.831 (high)
PL·min21 21.2 (poor) r 5 0.806 (high)
Sprint PL 23.4 (poor) r 5 0.949 (very high)
PL·min21 22.1 (poor) r 5 0.865 (high)
Barrett et al. (16) Catapult MinimaxX S4 (100 Hz) Two standardized incremental treadmill running PL 5.9 (moderate) ICC 5 0.93 (very high)
tests (7–16 km·h21) PL anterior- 9.1 (moderate) ICC 5 0.92 (very high)
posterior
PL medio-lateral 12.0 (poor) ICC 5 0.80 (high)
PL vertical 6.3 (moderate) ICC 5 0.93 (very high)
Beato et al. (23) STATSports Apex (100 Hz) Yo-Yo intermittent recovery level 1 Dynamic stress load ICC 5 0.681 (moderate)
Boyd et al. (27) Catapult MinimaxX 2.0 (100 Hz) Between device PL r 5 0.997 (very high)
Static 1.10 (good)
Hydraulic shaker at 0.5 g 1.04 (good)
Hydraulic shaker at 3.0 g 1.02 (good)
Australian football 1.94 (good)
Within device PL
Static 1.01 (good)
Hydraulic shaker at 0.5 g 0.91 (good)
Hydraulic shaker at 3.0 g 1.05 (good)
Clubb et al. (42) Catapult OptimEye S5 (100 Hz) 7 v 7 1 6 game PL ICC 5 0.79 (high)
PL·min21 ICC 5 0.79 (high)
PL·m21 ICC 5 0.79 (high)
PL anterior- ICC 5 0.39 (low)
posterior
PL mediolateral ICC 5 0.59 (moderate)
PL vertical ICC 5 0.56 (moderate)
10 v 10 game PL ICC 5 0.83 (high)
PL·min21 ICC 5 0.83 (high)
PL·m21 ICC 5 0.91 (very high)
PL anterior- ICC 5 0.66 (moderate)
posterior
PL mediolateral ICC 5 0.79 (high)
PL vertical ICC 5 0.80 (high)
11 v 11 game PL ICC 5 0.81 (high)
PL·min21 ICC 5 0.82 (high)
PL·m21 ICC 5 0.88 (high)
PL anterior- ICC 5 0.53 (moderate)
posterior
PL mediolateral ICC 5 0.76 (high)
PL vertical ICC 5 0.80 (high)
Gomez-Carmona et al. (80) RealTrack WIMU PRO (1000 Hz) Treadmill incremental running PL 3.52 (good) ICC 5 0.93 (very high)
Track IR PL 6.57 (moderate) ICC 5 0.88 (high)
Nicolella et al. (139) Catapult OptimEye S5 (100 Hz) Shaker table oscillating at 3.0 g
Anterior-posterior Catapult PL Median 0.05
(small)
Calculated PL Median 0.08
(small)
Mediolateral Catapult PL Median 0.06
(small)
Calculated PL Median 0.36
(small)
Vertical Catapult PL Median 0.06
(small)
Calculated PL Median 0.38
(small)
Wylde et al. (206) VX sport Log (100 Hz) Badminton-specific incremental test PL vertical 1.0 (good) ICC 5 0.996 (very high)
PL anterior- 2.3 (good) ICC 5 0.979 (very high)
posterior
PL mediolateral 2.3 (good) ICC 5 0.958 (very high)
*IR 5 incremental running; PL 5 PlayerLoad.
†Magnitude of correlations interpreted as 0.0 # negligible , 0.30; 0.3 # low , 0.50; 0.50 # moderate , 0.70; 0.70 # high , 0.90; 0.90 # very high # 1.0 (93,139). Coefficient of variation (CV)
interpreted as good # 5% , moderate # 10% , poor (173).

e9

Copyright © 2024 National Strength and Conditioning Association. Unauthorized reproduction of this article is prohibited.
 A Review of the Validity and Reliability of Accelerometers (2024) 00:00

acceleration sprints elicited significantly greater Dynamic Stress However, some studies have used a derivative of PlayerLoad by
Load (d 5 0.43, small) than self-selected submaximal acceleration isolating “PlayerLoadSlow” defined as the PlayerLoad at veloc-
intensities within the same repeated sprint protocol (20). This is ities ,0.2 m·s21. Roe et al. (160) reported moderate to high
despite no significant differences in peak speed, average speed, or correlations with collision count when using PlayerLoad Slow for
total distance between the 2 protocols, suggesting a potential rugby union backs (r 5 0.613 [moderate]) and forwards (r 5
association with external or internal forces and a capability to 0.701 [high]), improving the correlation compared with standard
distinguish these intensities. PlayerLoad in backs (r 5 0.477 [low]) and remaining high in
This is also supported by findings related to change of di- forwards (r 5 0.727 [high]).
rection, which includes both CoM deceleration and acceleration, Triaxial Body Load was found to be invalid for assessing
necessitating relatively high impulses and often associated with tackling impacts in rugby league due to high vertical accelerations
high peak GRF (173,199) and joint moments (25). Therefore, but was corrected by disregarding vertical data to generate a 2D
more intense or more frequent changes of direction would be Body Load metric. This suggested that large vertical accelerations
expected to result in greater accelerometer-based metric values. may mask the horizontal accelerations from collisions (50,132).
Indeed, cumulative acceleration magnitude increases with the Because this may be attributed to high magnitudes of vertical
number of changes of direction within shuttle run protocols of force in a running foot strike, it stands to reason that Play-
equal total distance (182). Similarly, PlayerLoad increases with erLoadSlow may improve the quality of information gathered
increased prescribed direction changes within a task matched for with vertical accelerations minimized (141,208). Conversely,
overall time and distance (97). The matching of overall time Hulin et al. (97) reported a lower correlation for PlayerLoadSlow
would have further augmented the acceleration and deceleration of r 5 0.47 (moderate) in rugby league, suggesting that although it
demands of the task beyond the isolated effects of the direction is promising, PlayerLoad and its derivatives should be used with
changes. Multiple studies have recognized that high-intensity caution when quantifying collision workloads, and consideration
actions over minimal distance, such as rapid accelerations and should be given to the activities being monitored.
decelerations, and changes of direction, can have a significant
contribution to an athlete’s load (5,28,54,86). These actions may Relationships With Internal Load Measures. There is a lack of
not be reflected in traditional distance-based GNSS metrics, such research investigating the relationship between accelerometer-
as distance covered (140,155) or running speed (94). Therefore, based load metrics and measures or estimates of internal bio-
accelerometer metrics with the potential to quantify these loads mechanical load, such as tissue-specific loads on the muscles,
could provide valuable insight into a significant contributor to bones, tendons, and ligaments of the musculoskeletal system. This
overall athlete load. is mostly due to the difficulty in obtaining field-based estimates of
these loading parameters (191). Recent methodological devel-
Running Speed. The positive relationship between running ac- opments offer the potential for future research and developments
celeration and peak running speed suggests that accelerometer- in this area, which will be discussed later in this review. Therefore,
based metrics may increase with running speed as with running this section focuses on studies investigating the relationship be-
acceleration (20,207). Indeed, overall and axis-specific Player- tween accelerometer-based load metrics and measures of internal
Load increased approximately linearly with treadmill running physiological load.
speed from 7 to 16 km·h21 (16). PlayerLoad is also greater in Overall, correlations with measures of internal physiological
repeated maximal sprints (mean peak speed: 6.9 6 0.3 m·s21) load have been mixed. Low and negligible correlations with
compared with intermittent incremental high-intensity running blood lactate measures suggest that it is unlikely accelerometer-
(6.2 6 0.3 m·s21) (78). These results were further supported by based load measures will coincide with biochemical blood
findings that peak instantaneous accelerations from GPSports markers (184), which is understandable given that lactate con-
units were greater during two 30-m sprints (8.28 6 1.81 g and centrations are unlikely to be primarily driven by mechanical
8.33 6 1.56 g) than two 10-m sprints (7.32 6 2.25 g and 7.03 6 load. However, there are moderate to very high correlations with
1.82 g) (196). It is also noteworthy, however, that the detection of heart rate measures (16,179) and TRIMP variations (149,167),
magnitudes above 5 g was more reliable during the 10-m sprint and moderate to high correlations with high-intensity in-
(CV: 4.7 vs 14.2%). The relationship between running speed and termittent running capacity (final velocity within 30–15 in-
accelerometer-based metric values could be attributed to in- termittent fitness test) (149) and YoYo tests for aerobic capacity
creased stride frequency and GRF in each axis at greater running (113,142). These can likely be attributed to players with greater
speeds. However, differences in running kinematics mean that aerobic capacity being capable of maintaining greater workloads
accelerometer-based metric results are not comparable between (166). There are some negligible to moderate correlations for
athletes and may vary with increasing fatigue (16). Body Load and session RPE variations (83,119,197); however,
there are moderate to high correlations with RPE and session RPE
Collisions and Impacts. Accurately quantifying collision loads is for PlayerLoad (34,66,75,131,167) and Dynamic Stress Load
a common challenge for practitioners in contact sports. Collisions (60,183). This demonstrates a contrast between PlayerLoad and
and impacts have been associated with significant increases in Dynamic Stress Load (moderate to high) and Body Load (negli-
heart rate, session RPE, creatine kinase levels, decreases in upper- gible to moderate), possibly because of differences in either the
body neuromuscular function, decreased perception of well- equations applied or the activities investigated. However, re-
being, increased fatigue, and possible injury risk (71,104,161). garding correlations with session RPE, it should be noted that
Although there is minimal evidence to evaluate the direct re- even a completely uncorrelated load·min21 and RPE may corre-
lationship between Dynamic Stress Load and collisions, Body late significantly once both are multiplied by the same session
Load has been positively evaluated (14) and applied (46,52) for duration (128). Although accelerometer loads cannot be perfectly
the purpose of quantifying collisions in contact sports. Studies correlated with internal physiological loads, there are examples of
have also identified strong associations between PlayerLoad and moderate to very high correlations across internal physiological
collisions for type, magnitude, and frequency (71,73,160). load measures that may support the use of accelerometer-based

e10

Copyright © 2024 National Strength and Conditioning Association. Unauthorized reproduction of this article is prohibited.
 A Review of the Validity and Reliability of Accelerometers (2024) 00:00 | www.nsca.com

load metrics as a tool for monitoring athlete load for specific monitored, as well as the intended application. The manu-
purposes. However, practitioners should ensure that they are facturers and metrics discussed in this review must be considered
aware of the capabilities specific to the manufacturer and algo- individually because of variations in the processing of acceler-
rithms used, as well as alternative methods for measuring physi- ometer signals before metric calculation and in the application of
ological loads directly or with greater construct validity. different manufacturer algorithms. Accelerometer-based metrics
can provide valid estimations of specific types of load under
Fatigue. Fatigue has been shown to influence running mechanics, specific circumstances but cannot directly assess internal or ex-
resulting in increases in peak vertical impact forces and peak ternal forces or mechanical load with absolute accuracy
propulsive forces (19,41). Accelerometer-based metrics could (47,61,162,189). Typically, raw data overestimates in-
provide valuable information if they can demonstrate or detect stantaneous accelerations, with error increasing with acceleration
the onset or accumulation of fatigue. It has been suggested that magnitude (61,196,204). In addition, upper back–mounted
PlayerLoad could reflect changes in mediolateral and ante- GNSS units are always susceptible to some incidental harness
roposterior plane movement quality resulting from fatigue (143). movement (61,132,206). These issues can be partially accounted
Training loads in a previous 2 and 7 day period, and so possible for by refined signal processing (e.g., appropriate filtering), an
fatigue, have significantly affected Dynamic Stress Load in Gaelic area requiring further study. Interdevice reliability results mean
football (48). In another study, absolute and relative (per minute) that practitioners should make comparisons within device where
Dynamic Stress Load were moderately affected by fatigue possible and plan unit use accordingly (15,16,42,47,80,139).
(i.e., differences between the same protocol performed before and Therefore, practitioners should consider the implications of the
after maximal repeated sprints). Therefore, it appears possible data collection method, type of activity, and surface on acceler-
that intraindividual differences in Dynamic Stress Load magni- ometer data. Testing and monitoring conditions should remain
tudes could be used to indicate acute fatigue (23,48). However, consistent where possible and avoid making between-athlete,
there are relatively few studies investigating this relationship, and manufacturer, or device comparisons. Finally, data processing
further research should be undertaken to establish causal pipelines should ideally include checks and countermeasures for
relationships. sensor limitations or possible errors (31,95).
Caution is urged when interpreting accelerations from 1 or
Injury. Although circumstance and mechanism of injury may several segments as a measure of external whole-body bio-
differ, it is ultimately a result of applied forces, stresses, and mechanical loading (191), as GNSS unit–derived accelerations
strains that cause damage exceeding the strength and repairability neglect the influence of other body segment accelerations, con-
of the tissue (12,62,107,108,130). As a result, mechanical loading siderable postimpact shock wave attenuation inferior to the sen-
is considered a key contributor to tissue damage accumulation sor (117,129), and the contribution of various frequency
and therefore athletic injuries (108). Knowledge of relationships components to the overall acceleration signal (129,174). Even an
between accelerometer-based metrics and injury risk would be entirely accurate measure of GRF or CoM acceleration
particularly useful. A 4-week accumulation of PlayerLoad (but (i.e., external biomechanical training load) may not correlate with
not 2- or 3-week accumulations) had a moderate association with the internal forces experienced by specific tissues (i.e., internal
injury probability in rugby league, estimated from injury occur- biomechanical training load) if muscle forces are not accounted
rence {Area-Under-Curve (95% confidence interval [CI]) 5 0.585 for within the causal framework (126). Combining intensity and
[0.482–0.672]; p 5 0.048} (51). In addition, Ehrmann et al. (63) duration to operationalize a single cumulative training load
reported that players had significantly lower Body Load 1 week metric also has its limitations (56,100,145,156,175). The accu-
(15.4% lower) and 4 weeks (9.0%) before injury when compared mulation of training load over time fails to consider the pattern of
with the season average. This was suggested to possibly reflect that load, for example, constant or variable, number and mag-
voluntary or involuntary reduced outputs before the injury due to nitude of individual loading cycles, rest periods, and the like.
fatigue or soreness. Thus, both increased and decreased Indeed, it may be better to consider intensity, duration, and the
accelerometer-based metric values have been associated with interaction between the 2 separately in an overall statistical model
greater injury risk through retrospective analysis, highlighting the (56,100,175) or even in a time-varying manner that reflects the
need for a clear causal framework and prospective investigations. true dynamic nature of many biological systems (100,178). Fi-
Furthermore, Wiig et al. (202) found PlayerLoad to be a poor nally, it should be noted that dose-response relationships between
predictor of muscle damage following competitive soccer. Despite accelerometer-based metrics and other aspects of the bio-
this, the literature overall suggests some potential to apply mechanical load-adaptation pathway (189) remain mostly
accelerometer-based metrics toward injury monitoring. Research unknown.
should establish further relationships between accelerometer-
based cumulative load metrics and injury risk, such as in different
environments, and considering other current accelerometer-based
Suggestions for Future Research
metrics, such as Dynamic Stress Load. It is likely that any such
efforts will need to focus on specific injuries as part of a clear For accelerometer-based metrics to effectively contribute to
causal framework and quantify or predict tissue loading cycle training prescription, injury risk reduction, or rehabilitation as
magnitudes and frequencies as well as intercycle recovery peri- part of a causal framework, it is necessary to quantify dose-
ods (108). response relationships with other aspects of the relevant bio-
mechanical load-adaptation pathways. This includes knowledge
Limitations of Global Navigation Satellite System–Housed of how to effectively manipulate accelerometer-based metrics
Accelerometers for Load Monitoring. Overall, the body of evi- through training design, as well as knowledge regarding the
dence has largely shown upper back–mounted accelerometers to relationships between these external biomechanical training load
be valid and reliable, but this is dependent on multiple factors, metrics and subsequent internal biomechanical training load,
such as the manufacturer, processing, filtering, and activity being tissue damage, and remodeling.

e11

Copyright © 2024 National Strength and Conditioning Association. Unauthorized reproduction of this article is prohibited.
 A Review of the Validity and Reliability of Accelerometers (2024) 00:00

One area of growing focus is the need and current relative

inability to measure internal biomechanical training loads outside Metric validity is dependent on the specific purpose, such as,
of a laboratory (191), against which to assess accelerometer- in rugby collisions where the larger vertical acceleration
based metrics. This is particularly relevant when seeking to as- component may mask accelerations in horizontal axes
sociate training load measures with injury likelihood or inform (50,55,132), the exclusion of vertical accelerations may im-
progressive overloading of the tissue during rehabilitation or in- prove the application. Although between-device reliability
jury risk reduction programs. Greater insight into internal bio- dictates that practitioners should minimize comparisons be-
mechanical training load can perhaps be gained when tween different units used by the same athlete, avoid between-
accelerometer or inertial measurement unit data are used along- athlete comparisons because of potential differences in mass
side existing models of tissue stress response or as inputs to and technique, and avoid comparisons between manu-
neuromusculoskeletal models (4,108,195). Open-source meth- facturers because of metric algorithm and processing varia-
odologies exist for the use of multiple IMUs as input to whole- tions (15,16,42,47,80,139). Practitioners can use the reported
body musculoskeletal computer simulation models that could validity and reliability values to inform decisions regarding
estimate internal biomechanical loads (4,177), but the number or selection and application of metrics, and the determination of
size of sensors required will likely need to be reduced before any magnitudes of change necessary to be considered “meaning-
real-world application for training load monitoring. However, ful” (i.e., greater than chance or random variation). An ex-
physics-based or machine learning models can be used to predict ample of this can be seen in Swinton et al. (180), which
accurate lower limb bone loads from minimal wearable technol- demonstrates the conversion of CV into CI to identify “true
ogy data (65,127). Alternatively, it may be possible to identify score” ranges.
algorithms to recognize certain activities (38,77,124,146) and
then scale tissue-specific load estimates from laboratory-based
estimates using an intensity measure derived from GNSS units. References
Any developed tools for the estimation of accurate external or 1. Agmon N, Ahituv N. Assessing data reliability in an information system.
internal biomechanical training loads may facilitate the pre- J Manag Inf Syst 4: 34–44, 1987.
2. Aguiar MVD, Botelho GMA, Gonçalves BSV, Sampaio JE. Physiological
scription and monitoring of specific loads, as well as the ranking responses and activity profiles of football small-sided games. J Strength
of exercises in terms of specific tissue loading (59,163), or the Cond Res 27: 1287–1294, 2013.
provision of haptic (57,76) or audio (29,187) biofeedback. Before 3. Akenhead R, Nassis GP. Training load and player monitoring in high-
evaluating new or existing metrics, it would be useful to quantify level football: Current practice and perceptions. Int J Sports Physiol
the desired level of validity or reliability (e.g., how accurate is Perform 11: 587–593, 2016.
4. Al Borno M, O’Day J, Ibarra V, et al. OpenSense: An open-source toolbox
accurate enough?). This will depend on the intended application. for inertial-measurement-unit-based measurement of lower extremity ki-
Finally, we should not only assess the metrics themselves but also nematics over long durations. J NeuroEng Rehabil 19: 22, 2022.
the consequences of prescribing changes in those metrics. 5. Alexander JP, Hopkinson TL, Wundersitz DWT, Serpell BG, Mara JK,
Ball NB. Validity of a wearable accelerometer device to measure average
acceleration values during high-speed running. J Strength Cond Res 30:
3007–3013, 2016.
Practical Applications 6. Alghannam AF. Metabolic limitations of performance and fatigue in
football. Asian J Sports Med 3: 65–73, 2012.
The available body of evidence demonstrates multiple poten- 7. Andersen MB, Williams JM. A model of stress and athletic injury: Pre-
tial effective applications for accelerometer-based load metrics diction and prevention. J Sport Exerc Psychol 10: 294–306, 1988.
in estimating and monitoring athlete load. However, 8. Armitage M, McErlain-Naylor SA, Devereux G, et al. On-field re-
habilitation in football: Current practice and perceptions. A survey of the
researchers and practitioners using this technology should
English premier league and football league. Sci Med Footb: 1–10, 2024.
always consider their choice of measurement device or metric 9. Aroganam G, Manivannan N, Harrison D. Review on wearable tech-
in relation to their specific objective. Practitioners applying nology sensors used in consumer sport applications. Sensors 19: 1983,
these metrics should remain aware of the limitations, recog- 2019.
nize what is known about the origin of these metrics, and 10. Atkinson G, Nevill AM. Statistical methods for assessing measurement
error (reliability) in variables relevant to sports medicine. Sports Med 26:
ensure results are processed, applied, and interpreted with 217–238, 1998.
these in mind. For example, GNSS-housed accelerometer 11. Aughey RJ. Applications of GPS technologies to field sports. Int J Sports
signals can be valid for estimating trunk accelerations when Physiol Perform 6: 295–310, 2011.
filtered appropriately; appropriate low-pass cutoff frequen- 12. Bahr R, Krosshaug T. Understanding injury mechanisms: A key com-
ponent of preventing injuries in sport. Br J Sports Med 39: 324–329,
cies may be approximately 20 Hz for high-impact activities
2005.
and 10 Hz for lower-impact activities. Upper back–mounted 13. Barbero-Alvarez JC, Soto VM, Barbero-Alvarez V, Granda-Vera J.
accelerometer signals can be affected by factors, such as in- Match analysis and heart rate of futsal players during competition.
cidental harness or attachment movement (61,132,206); J Sports Sci 26: 63–73, 2008.
therefore, sensor fixation should be as tight as is physically 14. Barr M, Beaver T, Turczyn D, Cornish S. Validity and reliability of 15 Hz
global positioning system units for assessing the activity profiles of uni-
comfortable to minimize additional displacement versity football players. J Strength Cond Res 33: 1371–1379, 2019.
(32,95,118,186). In addition, studies assessing reliability of 15. Barreira P, Robinson MA, Drust B, Nedergaard N, Raja Azidin RMF,
accelerometer-based load metrics found that reliability can be Vanrenterghem J. Mechanical Player Load using trunk-mounted accel-
reduced between devices, when compared between athletes, erometry in football: Is it a reliable, task- and player-specific observa-
tion? J Sports Sci 35: 1674–1681, 2017.
between activity types, and on different surfaces. As a result, it
16. Barrett S, Midgley A, Lovell R. PlayerLoad: Reliability, convergent val-
is recommended to maintain consistency in testing and mon- idity, and influence of unit position during treadmill running. Int J Sports
itoring conditions and allocate the same devices to the same Physiol Perform 9: 945–952, 2014.
athletes where possible. When interpreting accelerometer- 17. Bartlett JD, O’Connor F, Pitchford N, Torres-Ronda L, Robertson SJ.
based metrics, practitioners should regard the validity for Relationships between internal and external training load in team-sport
athletes: Evidence for an individualized approach. Int J Sports Physiol
purpose and relative error both within and between devices. Perform 12: 230–234, 2017.

e12

Copyright © 2024 National Strength and Conditioning Association. Unauthorized reproduction of this article is prohibited.
 A Review of the Validity and Reliability of Accelerometers (2024) 00:00 | www.nsca.com

18. Bataller-Cervero AV, Gutierrez H, DeRenterı́a J, et al. Validity and re- 43. Colby MJ, Dawson B, Heasman J, Rogalski B, Gabbett TJ. Accelerom-
liability of a 10 Hz GPS for assessing variable and mean running speed. eter and GPS-derived running loads and injury risk in elite Australian
J Hum Kinet 67: 17–24, 2019. footballers. J Strength Cond Res 28: 2244–2252, 2014.
19. Bazuelo-Ruiz B, Durá-Gil JV, Palomares N, Medina E, Llana-Belloch S. 44. Connor M, Beato M, O’Neill M. Adaptive athlete training plan gener-
Effect of fatigue and gender on kinematics and ground reaction forces ation: An intelligent control systems approach. J Sci Med Sport 25:
variables in recreational runners. PeerJ 6: e4489, 2018. 351–355, 2022.
20. Beato M, Drust B. Acceleration intensity is an important contributor to 45. Cordier S, Stewart PA. Exposure assessment. Handbook of Epidemiol-
the external and internal training load demands of repeated sprint ogy. Berlin, Heidelberg: Springer Berlin Heidelberg, 2005. pp. 437–462.
exercises in soccer players. Res Sports Med 29: 67–76, 2021. 46. Coughlan GF, Green BS, Pook PT, Toolan E, O’Connor SP. Physical
21. Beato M, Coratella G, Stiff A, Iacono AD. The validity and between-unit game demands in elite rugby union: A global positioning system analysis
variability of GNSS units (STATSports apex 10 and 18 Hz) for mea- and possible implications for rehabilitation. J Orthop Sports Phys Ther
suring distance and peak speed in team sports. Front Physiol 9: 1288, 41: 600–605, 2011.
2018. 47. Crang ZL, Duthie G, Cole MH, Weakley J, Hewitt A, Johnston RD. The
22. Beato M, Devereux G, Stiff A. Validity and reliability of global posi- validity and reliability of wearable microtechnology for intermittent
tioning system units (STATSports viper) for measuring distance and peak team sports: A systematic review. Sports Med 51: 549–565, 2020.
speed in sports. J Strength Cond Res 32: 2831–2837, 2018. 48. Cullen BD, McCarren AL, Malone S. Ecological validity of self-reported
23. Beato M, De Keijzer KL, Carty B, Connor M. Monitoring fatigue during wellness measures to assess pre-training and pre-competition pre-
intermittent exercise with accelerometer-derived metrics. Front Physiol paredness within elite Gaelic football. Sport Sci Health 17: 163–172,
10: 780, 2019. 2021.
24. Bergamini E, Melis M, Lentola A, Camomilla V. Estimate of trunk in- 49. Cummins C, Orr R, O’Connor H, West C. Global positioning systems
clination during fast movements by inertial sensing. In: 31st In- (GPS) and microtechnology sensors in team sports: A systematic review.
ternational Conference on Biomechanics in Sports. Shiang T-Y, Ho Sports Med 43: 1025–1042, 2013.
W-H, Chenfu Huang P, Tsai C-L, eds. Taipei, Taiwan: International 50. Cummins C, McLean B, Conlan G, Coutts A. Influence of drill type and
Society of Biomechanics in Sports (ISBS), 2013. unit fitting on accelerometer derived loads in rugby league training. J Sci
25. Besier TF, Lloyd DG, Cochrane JL, Ackland TR. External loading of the Med Sport 20: e117–e118, 2017.
knee joint during running and cutting maneuvers. Med Sci Sports Exerc 51. Cummins C, Welch M, Inkster B, et al. Modelling the relationships be-
33: 1168–1175, 2001. tween volume, intensity and injury-risk in professional rugby league
26. Bezodis NE, North JS, Razavet JL. Alterations to the orientation of the players. J Sci Med Sport 22: 653–660, 2019.
ground reaction force vector affect sprint acceleration performance in 52. Cunniffe B, Proctor W, Baker JS, Davies B. An evaluation of the physi-
team sports athletes. J Sports Sci 35: 1817–1824, 2017. ological demands of elite rugby union using global positioning system
27. Boyd LJ, Ball K, Aughey RJ. The reliability of MinimaxX accelerometers tracking software. J Strength Cond Res 23: 1195–1203, 2009.
for measuring physical activity in Australian football. Int J Sports Physiol 53. Cust EE, Sweeting AJ, Ball K, Robertson S. Classification of Australian
Perform 6: 311–321, 2011. football kick types in-situation via ankle-mounted inertial measurement
28. Boyd LJ, Ball K, Aughey RJ. Quantifying external load in Australian units. J Sports Sci 39: 1330–1338, 2020.
football matches and training using accelerometers. Int J Sports Physiol 54. Dalen T, Ingebrigtsen J, Ettema G, Hjelde GH, Wisløff U. Player load,
Perform 8: 44–51, 2013. acceleration, and deceleration during forty-five competitive matches of
29. Brake Mte, Stolwijk N, Staal B, Van Hooren B. Using beat frequency in elite soccer. J Strength Cond Res 30: 351–359, 2016.
music to adjust running cadence in recreational runners: A randomized 55. Davies MJ, Young W, Farrow D, Bahnert A. Comparison of agility
multiple baseline design. Eur J Sport Sci 23: 345–354, 2023. demands of small-sided games in elite Australian football. Int J Sports
30. Bredt SdGT, Chagas MH, Peixoto GH, Menzel HJ, de Andrade AGP. Physiol Perform 8: 139–147, 2013.
Understanding player load: Meanings and limitations. J Hum Kinet 71: 56. de Vocht F, Burstyn I, Sanguanchaiyakrit N. Rethinking cumulative
5–9, 2020. exposure in epidemiology, again. J Expo Sci Environ Epidemiol 25:
31. Brewer C, Dawson B, Heasman J, Stewart G, Cormack S. Movement 467–473, 2015.
pattern comparisons in elite (AFL) and sub-elite (WAFL) Australian 57. DeJong Lempke AF, Stephens SL, Fish PN, et al. Sensor-based gait
football games using GPS. J Sci Med Sport 13: 618–623, 2010. training to reduce contact time for runners with exercise-related lower leg
32. Camomilla V, Dumas R, Cappozzo A. Human movement analysis: The pain: A randomised controlled trial. BMJ Open Sport Exerc Med 8:
soft tissue artefact issue. J Biomech 62: 1–4, 2017. e001293, 2022.
33. Camomilla V, Bergamini E, Fantozzi S, Vannozzi G. Trends supporting 58. Dellaserra CL, Gao Y, Ransdell L. Use of integrated technology in team
the in-field use of wearable inertial sensors for sport performance eval- sports: A review of opportunities, challenges, and future directions for
uation: A systematic review. Sensors 18: 873, 2018. athletes. J Strength Cond Res 28: 556–573, 2014.
34. Casamichana D, Castellano J, Calleja-Gonzalez J, San Román J, Cas- 59. Demangeot Y, Whiteley R, Gremeaux V, Degache F. The load borne by
tagna C. Relationship between indicators of training load in soccer the achilles tendon during exercise: A systematic review of normative
players. J Strength Cond Res 27: 369–374, 2013. values. Scand J Med Sci Sports 33: 110–126, 2023.
35. Castillo ER, Lieberman DE. Shock attenuation in the human lumbar 60. Dowling J. Long-term Associations between Subjective Ratings of
spine during walking and running. J Exp Biol 221: jeb177949, 2018. Wellness and Exertion and Parameters of Training Load in Elite Soccer
36. Castillo D, Weston M, McLaren SJ, Cámara J, Yanci J. Relationships over Two Seasons. Liverpool, UK: Liverpool John Moores University,
between internal and external match-load indicators in soccer match 2019.
officials. Int J Sports Physiol Perform 12: 922–927, 2017. 61. Edwards S, White S, Humphreys S, Robergs R, O’Dwyer N. Caution
37. Caulfield B, Reginatto B, Slevin P. Not all sensors are created equal: A using data from triaxial accelerometers housed in player tracking units
framework for evaluating human performance measurement technolo- during running. J Sports Sci 37: 810–818, 2019.
gies. NPJ Digit Med 2: 7, 2019. 62. Edwards WB. Modeling overuse injuries in sport as a mechanical fatigue
38. Chambers R, Gabbett TJ, Cole MH, Beard A. The use of wearable phenomenon. Exerc Sport Sci Rev 46: 224–231, 2018.
microsensors to quantify sport-specific movements. Sports Med 45: 63. Ehrmann FE, Duncan CS, Sindhusake D, Franzsen WN, Greene DA. GPS
1065–1081, 2015. and injury prevention in professional soccer. J Strength Cond Res 30:
39. Checkoway H, Pearce NE, Kriebel D. Research Methods in Occupa- 360–367, 2016.
tional Epidemiology. New York, NY: Oxford University Press, 2004. 64. Eignor DR. The standards for educational and psychological testing. In:
40. Chen KY, Bassett DR. The technology of accelerometry-based activity APA Handbook of Testing and Assessment in Psychology, Vol. 1: Test
monitors: Current and future. Med Sci Sports Exerc 37: S490–S500, Theory and Testing and Assessment in Industrial and Organizational
2005. Psychology. Washington, WA: American Psychological Association,
41. Clansey AC, Hanlon M, Wallace ES, Lake MJ. Effects of fatigue on 2013. pp. 245–250.
running mechanics associated with tibial stress fracture risk. Med Sci 65. Elstub L, Nurse C, Grohowski L, Volgyesi P, Wolf D, Zelik KE. Tibial
Sports Exerc 44: 1917–1923, 2012. bone forces can be monitored using shoe-worn wearable sensors during
42. Clubb J, Towlson C, Barrett S. Measurement properties of external running. J Sports Sci 40: 1741–1749, 2022.
training load variables during standardised games in soccer: Implica- 66. Enes A, Oneda G, Alves DL, et al. Determinant factors of the match-
tions for training and monitoring strategies. PLoS One 17: e0262274, based internal load in elite soccer players. Res Q Exerc Sport 92: 63–70,
2022. 2021.

e13

Copyright © 2024 National Strength and Conditioning Association. Unauthorized reproduction of this article is prohibited.
 A Review of the Validity and Reliability of Accelerometers (2024) 00:00

67. Esteves L, Santos P, Lago-Penas C. Relationship between training load 92. Horsley BJ, Tofari PJ, Halson SL, et al. Does site matter? Impact of
indicators and training periodization during preseason in elite football inertial measurement unit placement on the validity and reliability of
goalkeepers. Hum Mov 2018: 89–97, 2018. stride variables during running: A systematic review and meta-analysis.
68. Evans SA, James DA, Rowlands D, Lee JB. Evaluation of accelerometer- Sports Med 51: 1449–1489, 2021.
derived data in the context of cycling cadence and saddle height changes 93. Howe ST, Aughey RJ, Hopkins WG, Stewart AM, Cavanagh BP.
in triathlon. Sensors 21: 871, 2021. Quantifying important differences in athlete movement during collision-
69. Fan Y. Dynamic principle of center of mass in human walking. Br Bio- based team sports: Accelerometers outperform Global Positioning Sys-
technol J 3: 524–544, 2013. tems. 2017 IEEE International Symposium on Inertial Sensors and
70. Fortune E, Morrow MMB, Kaufman KR. Assessment of gait kinetics Systems (INERTIAL). Kauai, HI: IEEE, 2017. pp. 1–4.
using triaxial accelerometers. J Appl Biomech 30: 668–674, 2014. 94. Huggins RA, Giersch GEW, Belval LN, et al. The validity and reliability
71. Gabbett T, Jenkins D, Abernethy B. Physical collisions and injury during of global positioning system units for measuring distance and velocity
professional rugby league skills training. J Sci Med Sport 13: 578–583, during linear and team sport simulated movements. J Strength Cond Res
2010. 34: 3070–3077, 2020.
72. Gabbett T, Sancho I, Dingenen B, Willy RW. When progressing training 95. Hughes GTG, Camomilla V, Vanwanseele B, Harrison AJ, Fong DTP,
loads, what are the considerations for healthy and injured athletes? Br J Bradshaw EJ. Novel technology in sports biomechanics: Some words of
Sports Med 55: 947–948, 2021. caution. Sports Biomech 23: 393–401, 2024.
73. Gabbett TJ. Relationship between accelerometer load, collisions, and 96. Hulin BT, Gabbett TJ, Kearney S, Corvo A. Physical demands of match
repeated high-intensity effort activity in rugby league players. J Strength play in successful and less-successful elite rugby league teams. Int J Sports
Cond Res 29: 3424–3431, 2015. Physiol Perform 10: 703–710, 2015.
74. Gabbett TJ. Debunking the myths about training load, injury and per- 97. Hulin BT, Gabbett TJ, Johnston RD, Jenkins DG. PlayerLoad variables:
formance: Empirical evidence, hot topics and recommendations for Sensitive to changes in direction and not related to collision workloads in
practitioners. Br J Sports Med 54: 58–66, 2020. rugby league match play. Int J Sports Physiol Perform 13: 1136–1142,
75. Gallo T, Cormack S, Gabbett T, Williams M, Lorenzen C. Character- 2018.
istics impacting on session rating of perceived exertion training load in 98. Impellizzeri FM, Rampinini E, Coutts AJ, Sassi A, Marcora SM. Use of
Australian footballers. J Sports Sci 33: 467–475, 2015. RPE-based training load in soccer. Med Sci Sports Exerc 36: 1042–1047,
76. Galofaro E, D’Antonio E, Lotti N, Masia L. Rendering immersive haptic 2004.
force feedback via neuromuscular electrical stimulation. Sensors 22: 99. Impellizzeri FM, Rampinini E, Marcora SM. Physiological assessment of
5069, 2022. aerobic training in soccer. J Sports Sci 23: 583–592, 2005.
77. Ghosh I, Ramasamy Ramamurthy S, Chakma A, Roy N. DeCoach: Deep 100. Impellizzeri FM, Shrier I, McLaren SJ, et al. Understanding training load
learning-based coaching for badminton player assessment. Pervasive as exposure and dose. Sports Med 53: 1667–1679, 2023.
Mob Comput 83: 101608, 2022. 101. Jackson BM, Polglaze T, Dawson B, King T, Peeling P. Comparing global
78. Gibson NV, Henning G, Twist C. Movement characteristics, physio- positioning system and global navigation satellite system measures of
logical and perceptual responses of elite standard youth football players team-sport movements. Int J Sports Physiol Perform 13: 1005–1010,
to different high intensity running drills. Sci Med Footb 2: 281–287, 2018.
2018. 102. Jeffries AC, Marcora SM, Coutts AJ, Wallace L, McCall A, Impellizzeri
79. Gómez-Carmona CD, Bastida-Castillo A, Garcı́a-Rubio J, Ibáñez SJ, FM. Development of a revised conceptual framework of physical train-
Pino-Ortega J. Static and dynamic reliability of WIMU PRO accel- ing for use in research and practice. Sports Med 52: 709–724, 2022.
erometers according to anatomical placement. Proc Inst Mech Eng P J 103. John D, Miller R, Kozey-Keadle S, Caldwell G, Freedson P. Bio-
Sports Eng Technol 233: 238–248, 2019. mechanical examination of the “plateau phenomenon” in ActiGraph
80. Gómez-Carmona CD, Bastida-Castillo A, González-Custodio A, Olcina vertical activity counts. Physiol Meas 33: 219–230, 2012.
G, Pino-Ortega J. Using an inertial device (WIMU PRO) to quantify 104. Johnston RD, Gabbett TJ, Seibold AJ, Jenkins DG. Influence of physical
neuromuscular load in running: Reliability, convergent validity, and contact on neuromuscular fatigue and markers of muscle damage fol-
influence of type of surface and device location. J Strength Cond Res 34: lowing small-sided games. J Sci Med Sport 17: 535–540, 2014.
365–373, 2020. 105. Johnston RJ, Watsford ML, Austin DJ, Pine MJ, Spurrs RW. Movement
81. Gómez-Carmona CD, Bastida-Castillo A, Ibáñez SJ, Pino-Ortega J. profiles, match events, and performance in Australian football.
Accelerometry as a method for external workload monitoring in invasion J Strength Cond Res 30: 2129–2137, 2016.
team sports. A systematic review. PLoS One 15: e0236643, 2020. 106. Johnston RD, Weaving D, Hulin BT, Till K, Jones B, Duthie G. Peak
82. Gómez-Carmona CD, Pino-Ortega J, Rico-González M. Micro- movement and collision demands of professional rugby league compe-
electromechanical systems. In: The Use of Applied Technology in Team tition. J Sports Sci 37: 2144–2151, 2019.
Sport. Pino-Ortega J, Rico-González M, eds. New York, NY: Routledge, 107. Kalkhoven JT, Watsford ML, Impellizzeri FM. A conceptual model and
2021. pp. 50–71. detailed framework for stress-related, strain-related, and overuse athletic
83. Gomez-Piriz PT, Jiménez-Reyes P, Ruiz-Ruiz C. Relation between total injury. J Sci Med Sport 23: 726–734, 2020.
body load and session rating of perceived exertion in professional soccer 108. Kalkhoven JT, Watsford ML, Coutts AJ, Edwards WB, Impellizzeri FM.
players. J Strength Cond Res 25: 2100–2103, 2011. Training load and injury: Causal pathways and future directions. Sports
84. Hamill J, Derrick TR, Holt KG. Shock attenuation and stride frequency Med 51: 1137–1150, 2021.
during running. Hum Mov Sci 14: 45–60, 1995. 109. Kelly SJ, Murphy AJ, Watsford ML, Austin D, Rennie M. Reliability and
85. Harper DJ, Kiely J. Damaging nature of decelerations: Do we adequately validity of sports accelerometers during static and dynamic testing. Int J
prepare players? BMJ Open Sport Exerc Med 4: e000379, 2018. Sports Physiol Perform 10: 106–111, 2015.
86. Harper DJ, Carling C, Kiely J. High-intensity acceleration and de- 110. Kobsar D, Osis ST, Hettinga BA, Ferber R. Classification accuracy of
celeration demands in elite team sports competitive match play: A sys- a single tri-axial accelerometer for training background and experience
tematic review and meta-analysis of observational studies. Sports Med level in runners. J Biomech 47: 2508–2511, 2014.
49: 1923–1947, 2019. 111. Kokshenev VB. Dynamics of human walking at steady speeds. Phys Rev
87. Harrison AJ, McErlain-Naylor SA, Bradshaw EJ, et al. Recom- Lett 93: 208101, 2004.
mendations for statistical analysis involving null hypothesis significance 112. Koo TK, Li MY. A guideline of selecting and reporting intraclass cor-
testing. Sports Biomech 19: 561–568, 2020. relation coefficients for reliability research. J Chiropr Med 15: 155–163,
88. Hausler J, Halaki M, Orr R. Application of global positioning system 2016.
and microsensor technology in competitive rugby league match-play: A 113. Lee M, Mukherjee S. Relationship of training load with high-intensity
systematic review and meta-analysis. Sports Med 46: 559–588, 2016. running in professional soccer players. Int J Sports Med 40: 336–343,
89. Hinkle DE, Wiersma W, Jurs SG. Applied Statistics for the Behavioral 2019.
Sciences (5th ed.). Boston, MA: Houghton Mifflin, 2003. 114. Lee TA, Pickard AS. Exposure definition and measurement. In: De-
90. Ho C-S, Chang C-H, Hsu Y-J, et al. Feasibility of the energy expenditure veloping a Protocol for Observational Comparative Effectiveness Re-
prediction for athletes and non-athletes from ankle-mounted acceler- search: A User’s Guide. Velentgas P, Dreyer N, Nourjah P, eds.
ometer and heart rate monitor. Sci Rep 10: 8816, 2020. Rockville, MD: Agency for Healthcare Research and Quality (US), 2013.
91. Hollville E, Couturier A, Guilhem G, Rabita G. A novel accelerometry- 115. Lee JB, Sutter KJ, Askew CD, Burkett BJ. Identifying symmetry in run-
based metric to improve estimation of whole-body mechanical load. ning gait using a single inertial sensor. J Sci Med Sport 13: 559–563,
Sensors 21: 3398, 2021. 2010.

e14

Copyright © 2024 National Strength and Conditioning Association. Unauthorized reproduction of this article is prohibited.
 A Review of the Validity and Reliability of Accelerometers (2024) 00:00 | www.nsca.com

116. Leite WSS. Physiological demands in football, futsal and beach soccer: A 141. Nilsson J, Thorstensson A. Ground reaction forces at different speeds of
brief review. Eur J Phys Educ Sport Sci 2: 1–9, 2016. human walking and running. Acta Physiol Scand 136: 217–227, 1989.
117. Liew BXW, Zhu X, Zhai X, McErlain-Naylor SA, McManus C. Asso- 142. Owen AL, Newton M, Shovlin A, Malone S. The use of small-sided
ciation between fat and fat-free body mass indices on shock attenuation games as an aerobic fitness assessment supplement within elite level
during running. J Biomech 165: 112025, 2024. professional soccer. J Hum Kinet 71: 243–253, 2020.
118. Liu T, Inoue Y, Shibata K. Measurement of soft tissue deformation to 143. Page RM, Marrin K, Brogden CM, Greig M. Biomechanical and phys-
improve the accuracy of a body-mounted motion sensor. J Med Dev iological response to a contemporary soccer match-play simulation.
Trans ASME 3, 2009. J Strength Cond Res 29: 2860–2866, 2015.
119. Lovell TWJ, Sirotic AC, Impellizzeri FM, Coutts AJ. Factors affecting 144. Paquette MR, Miller RH. Reconciling new with old injury paradigms
perception of effort (session rating of perceived exertion) during rugby and the need to dig deeper—comment on Nigg et al. Curr Issues Sport Sci
league training. Int J Sports Physiol Perform 8: 62–69, 2013. (CISS), 2018.
120. Luczak T, Burch R, Lewis E, Chander H, Ball J. State-of-the-art review of 145. Passfield L, Murias JM, Sacchetti M, Nicolò A. Validity of the training-
athletic wearable technology: What 113 strength and conditioning load concept. Int J Sports Physiol Perform 17: 507–514, 2022.
coaches and athletic trainers from the USA said about technology in 146. Perri T, Reid M, Murphy A, Howle K, Duffield R. Validating an algo-
sports. Int J Sports Sci Coach 15: 26–40, 2020. rithm from a trunk-mounted wearable sensor for detecting stroke events
121. Lutz J, Memmert D, Raabe D, Dornberger R, Donath L. Wearables for in tennis. J Sports Sci 40: 1168–1174, 2022.
integrative performance and tactic analyses: Opportunities, challenges, 147. Pino-Ortega J, Oliva-Lozano JM, Gantois P, Nakamura FY, Rico-
and future directions. Int J Environ Res Public Health 17: 59, 2019. González M. Comparison of the validity and reliability of local posi-
122. Macadam P, Cronin J, Neville J, Diewald S. Quantification of the validity tioning systems against other tracking technologies in team sport: A
and reliability of sprint performance metrics computed using inertial systematic review. Proc Inst Mech Eng P J Sports Eng Technol 236:
sensors: A systematic review. Gait Posture 73: 26–38, 2019. 73–82, 2022.
123. Malone JJ, Lovell R, Varley MC, Coutts AJ. Unpacking the black box: 148. Polglaze T, Dawson B, Hiscock DJ, Peeling P. A comparative analysis of
Applications and considerations for using GPS devices in sport. Int J accelerometer and time–motion data in elite men’s hockey training and
Sports Physiol Perform 12: S218–S226, 2017. competition. Int J Sports Physiol Perform 10: 446–451, 2015.
124. Mannini A, Sabatini AM. Machine learning methods for classifying 149. Rabbani A, Kargarfard M, Castagna C, Clemente FM, Twist C.
human physical activity from on-body accelerometers. Sensors 10: Associations between selected training-stress measures and fitness
1154–1175, 2010. changes in male soccer players. Int J Sports Physiol Perform 14:
125. Mannini A, Intille SS, Rosenberger M, Sabatini AM, Haskell W. Activity 1050–1057, 2019.
recognition using a single accelerometer placed at the wrist or ankle. Med 150. Rabita G, Dorel S, Slawinski J, et al. Sprint mechanics in world-class
Sci Sports Exerc 45: 2193–2203, 2013. athletes: A new insight into the limits of human locomotion. Scand J Med
126. Matijevich ES, Branscombe LM, Scott LR, Zelik KE. Ground reaction Sci Sports 25: 583–594, 2015.
force metrics are not strongly correlated with tibial bone load when 151. Radzimiński Ł, Szwarc A, Padrón-Cabo A, Jastrze˛ bski Z. Correlations
running across speeds and slopes: Implications for science, sport and between body composition, aerobic capacity, speed and distance covered
wearable tech. PLoS One 14: e0210000, 2019. among professional soccer players during official matches. J Sports Med
127. Matijevich ES, Scott LR, Volgyesi P, Derry KH, Zelik KE. Combining Phys Fitness 60: 257–262, 2020.
wearable sensor signals, machine learning and biomechanics to estimate tibial 152. Rago V, Brito J, Figueiredo P, Krustrup P, Rebelo A. Relationship be-
bone force and damage during running. Hum Mov Sci 74: 102690, 2020. tween external load and perceptual responses to training in professional
128. McErlain-Naylor SA, Beato M. Factors influencing the jump football: Effects of quantification method. Sports 7: 68, 2019.
momentum—sprint momentum correlation: A data simulation. Eur J 153. Rago V, Brito J, Figueiredo P, et al. Methods to collect and interpret
Sport Sci 22: 1847–1855, 2022. external training load using microtechnology incorporating GPS in
129. McErlain-Naylor SA, King MA, Allen SJ. Surface acceleration trans- professional football: A systematic review. Res Sports Med 28: 437–458,
mission during drop landings in humans. J Biomech 118: 110269, 2021. 2020.
130. McIntosh AS. Risk compensation, motivation, injuries, and bio- 154. Randers MB, Nielsen JJ, Bangsbo J, Krustrup P. Physiological response
mechanics in competitive sport. Br J Sports Med 39: 2–3, 2005. and activity profile in recreational small-sided football: No effect of the
131. McLaren SJ, Macpherson TW, Coutts AJ, Hurst C, Spears IR, Weston number of players. Scand J Med Sci Sports 24: 130–137, 2014.
M. The relationships between internal and external measures of training 155. Rawstorn JC, Maddison R, Ali A, Foskett A, Gant N. Rapid directional
load and intensity in team sports: A meta-analysis. Sports Med 48: change degrades GPS distance measurement validity during intermittent
641–658, 2018. intensity running. PLoS One 9: e93693, 2014.
132. McLean BD, Cummins C, Conlan G, Duthie G, Coutts AJ. The fit 156. Renfree A, Casado A, McLaren S. Re-Thinking athlete training loads:
matters: Influence of accelerometer fitting and training drill demands on Would you rather have one big rock or lots of little rocks dropped on
load measures in rugby league players. Int J Sports Physiol Perform 13: your foot? Res Sports Med 30: 573–576, 2022.
1083–1089, 2018. 157. Reynolds J, Connor M, Jamil M, Beato M. Quantifying and comparing
133. Milioni F, Vieira LHP, Barbieri RA, et al. Futsal match-related fatigue the match demands of U18, U23, and 1ST team English professional
affects running performance and neuromuscular parameters but not soccer players. Front Physiol 12: 706451, 2021.
finishing kick speed or accuracy. Front Physiol 7: 2016, 2016. 158. Ribeiro JN, Gonçalves B, Coutinho D, Brito J, Sampaio J, Travassos B.
134. Moe-Nilssen R, Helbostad JL. Estimation of gait cycle characteristics by Activity profile and physical performance of match play in elite futsal
trunk accelerometry. J Biomech 37: 121–126, 2004. players. Front Psychol 11: 1709, 2020.
135. Mukaka MM. Statistics corner: A guide to appropriate use of correlation 159. Rispens SM, Cox LGE, Ejupi A, Delbaere K, Annegarn J, Bonomi AG.
coefficient in medical research. Malawi Med J 24: 69–71, 2012. Validation of walking speed estimation from trunk mounted accel-
136. Nagahara R, Mizutani M, Matsuo A, Kanehisa H, Fukunaga T. Asso- erometers for a range of walking speeds. Sensors 21: 1854, 2021.
ciation of sprint performance with ground reaction forces during accel- 160. Roe G, Halkier M, Beggs C, Till K, Jones B. The use of accelerometers to
eration and maximal speed phases in a single sprint. J Appl Biomech 34: quantify collisions and running demands of rugby union match-play. Int
104–110, 2018. J Perform Anal Sport 16: 590–601, 2016.
137. Nedergaard NJ, Robinson MA, Eusterwiemann E, Drust B, Lisboa PJ, 161. Roe G, Darrall-Jones J, Till K, et al. The effect of physical contact on
Vanrenterghem J. The relationship between whole-body external load- changes in fatigue markers following rugby union field-based training.
ing and body-worn accelerometry during team-sport movements. Int J Eur J Sport Sci 17: 647–655, 2017.
Sports Physiol Perform 12: 18–26, 2017. 162. Roell M, Mahler H, Lienhard J, Gehring D, Gollhofer A, Roecker K.
138. Nedergaard NJ, Verheul J, Drust B, et al. The feasibility of predicting Validation of wearable sensors during team sport-specific movements in
ground reaction forces during running from a trunk accelerometry indoor environments. Sensors 19: 3458, 2019.
driven mass-spring-damper model. PeerJ 6: e6105, 2018. 163. Sancho I, Willy RW, Morrissey D, Malliaras P, Lascurain-Aguirrebeña I.
139. Nicolella DP, Torres-Ronda L, Saylor KJ, Schelling X. Validity and re- Achilles tendon forces and pain during common rehabilitation exercises
liability of an accelerometer-based player tracking device. PLoS One 13: in male runners with Achilles tendinopathy. A laboratory study. Phys
e0191823, 2018. Ther Sport 60: 26–33, 2023.
140. Nikolaidis PT, Clemente FM, van der Linden CMI, Rosemann T, Knechtle 164. Santos PJ, Valente AP, Soares JMC. Aerobic capacity versus total dis-
B. Validity and reliability of 10-hz global positioning system to assess in- tance covered during a game in elite soccer players. Med Sci Sports Exerc
line movement and change of direction. Front Physiol 9: 228, 2018. 33: S157, 2001.

e15

Copyright © 2024 National Strength and Conditioning Association. Unauthorized reproduction of this article is prohibited.
 A Review of the Validity and Reliability of Accelerometers (2024) 00:00

165. Schelling X, Torres L. Accelerometer load profiles for basketball-specific 189. Vanrenterghem J, Nedergaard NJ, Robinson MA, Drust B. Training load
drills in elite players. J Sports Sci Med 15: 585–591, 2016. monitoring in team sports: A novel framework separating physiological
166. Schmitz B, Pfeifer C, Kreitz K, Borowski M, Faldum A, Brand S-M. The and biomechanical load-adaptation pathways. Sports Med 47:
Yo-Yo intermittent tests: A systematic review and structured compen- 2135–2142, 2017.
dium of test results. Front Physiol 9: 870, 2018. 190. Verheul J, Gregson W, Lisboa P, Vanrenterghem J, Robinson MA.
167. Scott BR, Lockie RG, Knight TJ, Clark AC, Janse de Jonge XAK. A Whole-body biomechanical load in running-based sports: The validity of
comparison of methods to quantify the in-season training load of pro- estimating ground reaction forces from segmental accelerations. J Sci
fessional soccer players. Int J Sports Physiol Perform 8: 195–202, 2013. Med Sport 22: 716–722, 2019.
168. Scott MTU, Scott TJ, Kelly VG. The validity and reliability of global 191. Verheul J, Nedergaard NJ, Vanrenterghem J, Robinson MA. Measuring
positioning systems in team sport: A brief review. J Strength Cond Res biomechanical loads in team sports—from lab to field. Sci Med Footb 4:
30: 1470–1490, 2016. 246–252, 2020.
169. Scott PJ, Brown AW, Adedeji T, et al. A review of measurement practice 192. Vickery WM, Dascombe BJ, Baker JD, Higham DG, Spratford WA,
in studies of clinical decision support systems 1998–2017. J Am Med Duffield R. Accuracy and reliability of GPS devices for measurement of
Inform Assoc 26: 1120–1128, 2019. sports-specific movement patterns related to cricket, tennis, and field-
170. Sechrest L. Validity of measures is No simple matter. Health Serv Res 40: based team sports. J Strength Cond Res 28: 1697–1705, 2014.
1584–1604, 2005. 193. Waegli A, Skaloud J. Optimization of two GPS/MEMS-IMU integration
171. Serpiello FR, Hopkins WG, Barnes S, et al. Validity of an ultra-wideband strategies with application to sports. GPS Solut 13: 315–326, 2009.
local positioning system to measure locomotion in indoor sports. J Sports 194. Wakeling JM, Nigg BM, Rozitis AI. Muscle activity damps the soft tissue
Sci 36: 1727–1733, 2018. resonance that occurs in response to pulsed and continuous vibrations.
172. Shechtman O. The coefficient of variation as an index of measurement J Appl Physiol 93: 1093–1103, 2002.
reliability. In: Methods of Clinical Epidemiology. Heidelberg, Germany: 195. Wakeling JM, Febrer-Nafrı́a M, De Groote F. A review of the efforts to
Springer Berlin, 2013. pp. 39–49. develop muscle and musculoskeletal models for biomechanics in the last
173. Shimokochi Y, Ide D, Kokubu M, Nakaoji T. Relationships among 50 years. J Biomech 155: 111657, 2023.
performance of lateral cutting maneuver from lateral sliding and hip 196. Waldron M, Worsfold P, Twist C, Lamb K. Concurrent validity and
extension and abduction motions, ground reaction force, and body test–retest reliability of a global positioning system (GPS) and timing
center of mass height. J Strength Cond Res 27: 1851–1860, 2013. gates to assess sprint performance variables. J Sports Sci 29: 1613–1619,
174. Shorten MR, Winslow DS. Spectral analysis of impact shock during 2011.
running. Int J Sport Biomech 8: 288–304, 1992. 197. Weaving D, Marshall P, Earle K, Nevill A, Abt G. Combining internal-
175. Shrier I, Wang C, Stokes T, Trejovargas J, Stovitz SD, Steele RJ. Causal and external-training-load measures in professional rugby league. Int J
effects, workload and injury risk: The importance of specifying the re- Sports Physiol Perform 9: 905–912, 2014.
search question. J Sci Med Sport 25: 574–578, 2022. 198. Weaving D, Whitehead S, Till K, Jones B. Validity of real-time data
176. Simons C, Bradshaw EJ. Do accelerometers mounted on the back pro- generated by a wearable microtechnology device. J Strength Cond Res
vide a good estimate of impact loads in jumping and landing tasks? 31: 2876–2879, 2017.
Sports Biomech 15: 76–88, 2016. 199. Welinski ML, Lee LN, McBroom B, Mufarreh B, Gidley AD. Ground
177. Slade P, Habib A, Hicks JL, Delp SL. An open-source and wearable reaction forces and temporal characteristics define cutting performance.
system for measuring 3D human motion in real-time. IEEE Trans Int J Exerc Sci 14: 211–221, 2021.
Biomed Eng 69: 678–688, 2022. 200. Wellman AD, Coad SC, Goulet GC, McLellan CP. Quantification of
178. Smith TJ, Kriebel D. Biologically based exposure assessment for epide- competitive game demands of NCAA division I college football players
miology. In: Current Topics in Occupational Epidemiology. Croydon, using global positioning systems. J Strength Cond Res 30: 11–19, 2016.
Unite Kingdom: Oxford University Press, 2013. pp. 199–217. 201. White E, Armstrong BK, Saracci R. Principles of Exposure Measurement
179. Sparks M, Coetzee B, Gabbett TJ. Internal and external match loads of in Epidemiology. Oxford, UK: Oxford University Press, 2008.
university-level soccer players: A comparison between methods. 202. Wiig H, Raastad T, Luteberget LS, Ims I, Spencer M. External load
J Strength Cond Res 31: 1072–1077, 2017. variables affect recovery markers up to 72 h after semiprofessional
180. Swinton PA, Hemingway BS, Saunders B, Gualano B, Dolan E. A sta- football matches. Front Physiol 10: 689, 2019.
tistical framework to interpret individual response to intervention: 203. Wundersitz DWT, Netto KJ, Aisbett B, Gastin PB. Validity of an upper-
Paving the way for personalized nutrition and exercise prescription. body-mounted accelerometer to measure peak vertical and resultant
Front Nutr 5: 41, 2018. force during running and change-of-direction tasks. Sports Biomech 12:
181. Tallent J, Higgins M, Parker N, et al. Quantification of bowling workload 403–412, 2013.
and changes in cognitive function in elite fast bowlers in training compared 204. Wundersitz DWT, Gastin PB, Richter C, Robertson SJ, Netto KJ. Val-
with Twenty20 Cricket. J Sports Med Phys Fitness 59: 35–41, 2019. idity of a trunk-mounted accelerometer to assess peak accelerations
182. Tang R, Murtagh C, Warrington G, et al. Directional change mediates during walking, jogging and running. Eur J Sport Sci 15: 382–390, 2015.
the physiological response to high-intensity shuttle running in pro- 205. Wundersitz DWT, Gastin PB, Robertson SJ, Netto KJ. Validity of
fessional soccer players. Sports 6: 39, 2018. a trunk-mounted accelerometer to measure physical collisions in contact
183. Tang R. The Influence of the Frequency of Directional Change on the sports. Int J Sports Physiol Perform 10: 681–686, 2015.
Physiological Responses to Intermittent Exercise in Elite Football 206. Wylde MJ, Lee MBC, Chee Yong L, Callaway AJ. Reliability and validity
Players. Liverpool, UK: Liverpool John Moores University, 2015. of GPS-embedded accelerometers forthe measurement of badminton
184. Taylor RJ, Sanders D, Myers T, Abt G, Taylor CA, Akubat I. The dose- specific player load. J Trainology 7: 34–37, 2018.
response relationship between training load and aerobic fitness in academy 207. Yıldız S, Ateş O, Gelen E, et al. The relationship between start speed,
rugby union players. Int J Sports Physiol Perform 13: 163–169, 2018. acceleration and speed performances in soccer. Universal J Educ Res 6:
185. Theodoropoulos JS, Bettle J, Kosy JD. The use of GPS and inertial devices 1697–1700, 2018.
for player monitoring in team sports: A review of current and future 208. Yu L, Mei Q, Xiang L, et al. Principal component analysis of the running
applications. Orthop Rev 12: 7863, 2020. ground reaction forces with different speeds. Front Bioeng Biotechnol 9:
186. Valiant GA, McMahon TA, Frederick E. In: A New Test to Evaluate the 629809, 2021.
Cushioning Properties of Athletic Shoes. Biomechanics X-B, Jonsson B, 209. Zadpoor AA, Nikooyan AA. The relationship between lower-extremity
eds. Champaign, IL: Human Kinetics, 1987. pp. 937–941. stress fractures and the ground reaction force: A systematic review. Clin
187. Van den Berghe P, Derie R, Bauwens P, et al. Reducing the peak tibial Biomech 26: 23–28, 2011.
acceleration of running by music‐based biofeedback: A quasi‐random- 210. Zandbergen MA, Ter Wengel XJ, van Middelaar RP, Buurke JH, Veltink
ized controlled trial. Scand J Med Sci Sports 32: 698–709, 2022. PH, Reenalda J. Peak tibial acceleration should not be used as indicator
188. van der Worp H, Vrielink JW, Bredeweg SW. Do runners who suffer of tibial bone loading during running. Sports Biomech: 1–18, 2023.
injuries have higher vertical ground reaction forces than those who re- 211. Zurutuza U, Castellano J, Echeazarra I, Casamichana D. Absolute and
main injury-free? A systematic review and meta-analysis. Br J Sports Med relative training load and its relation to fatigue in football. Front Psychol
50: 450–457, 2016. 8: 878, 2017.

e16

Copyright © 2024 National Strength and Conditioning Association. Unauthorized reproduction of this article is prohibited.