Accuracy of GPS sport watches in measuring distance in an ultramarathon running race

Abstract

Purpose
The aim of the study was to determine the accuracy of various global positioning system (GPS) sport watches in measuring distance throughout a 56 km running race.

Methods
The measured distance between timing mats was compared to the reported distance of GPS devices at the 2017 Two Oceans Marathon. Runners (n = 255) were divided into eight different categories based on GPS sport watch brand and model. The difference between distance measured by GPS and race markers was represented in metres (m) and as a relative error (%).

Results
The Garmin Fenix and cell phone categories had higher errors in measuring distance from the 16 km to finish (56 km) point compared to all other devices, except for activity watches.

Conclusions
The GPS sport watches in this study have an accuracy of 0.6 ± 0.3% to 1.9 ± 1.5% (median ± interquartile range) in reporting distance covered. This indicates that GPS sport watches are a valid and feasible method for sport scientists and coaches to measure performance and track training load. However, the small error associated with each brand needs to be considered when data are interpreted.

Full text

Original research
Accuracy of GPS sport watches
in measuring distance in an
ultramarathon running race
Rebecca E Johansson , Steffen T Adolph, Jeroen Swart and
Mike I Lambert
Abstract
Purpose: The aim of the study was to determine the accuracy of various global positioning system (GPS) sport watches
in measuring distance throughout a 56 km running race.
Methods: The measured distance between timing mats was compared to the reported distance of GPS devices at the
2017 Two Oceans Marathon. Runners (n ¼255) were divided into eight different categories based on GPS sport watch
brand and model. The difference between distance measured by GPS and race markers was represented in metres (m)
and as a relative error (%).
Results: The Garmin Fenix and cell phone categories had higher errors in measuring distance from the 16 km to finish
(56 km) point compared to all other devices, except for activity watches.
Conclusions: The GPS sport watches in this study have an accuracy of 0.6 0.3% to 1.9 1.5% (median interquartile
range) in reporting distance covered. This indicates that GPS sport watches are a valid and feasible method for sport
scientists and coaches to measure performance and track training load. However, the small error associated with each
brand needs to be considered when data are interpreted.
Keywords
Devices, endurance, technology, training, validity
Introduction
Wearable technology has, in the last decade, become
increasingly popular with endurance runners.
Wearables currently represent a $6-billion industry
1
with a projected $25-billion industry by as early as
2019.
2
A large category of wearables give information
about distance and speed travelled using Global
Navigation Satellite Systems such as GPS.
3,4
To obtain the best possible GPS readings, a high-
sampling frequency, open areas free from obstructions
such as tall buildings and clear skies are required.
3,5
Endurance runners have several GPS sport watch
brands to choose from such as GarminV
R, PolarV
R,
SuuntoV
Rand TomTomV
R. Within these brands, there
are several existing models, e.g. Garmin ForerunnerV
R
620, Suunto AmbitV
R3. It is important to evaluate
the reliability and accuracy of these GPS sport watches
for coaches and sport scientists to accurately measure
performance and track training load. To date,
most research regarding the accuracy of GPS devices
is related to team sports such as rugby and football,
and primarily focus on the accuracy in measuring high
intensity changes in speed.
6–11
The GPS devices used in
team sports are unique in their specifications and
cannot be directly compared to GPS sport watches
used in endurance running. There has also been
research regarding the validity and reliability of
consumer-wearable devices to measure heart rate,
12
Reviewer: Ranel Venter (Stellenbosch University, South Africa)
Division of Exercise Science & Sports Medicine, Department of Human
Biology, University of Cape Town, Sports Science Institute of South
Africa, Cape Town, South Africa
Corresponding author:
Rebecca E Johansson, Division of Exercise Science & Sports Medicine,
Department of Human Biology, University of Cape Town, Sports Science
Institute of South Africa, Boundary Road, Newlands, Cape Town 7725,
South Africa.
Email: JHNREB006@myuct.ac.za
International Journal of Sports Science
& Coaching
0(0) 1–8
!The Author(s) 2020
Article reuse guidelines:
sagepub.com/journals-permissions
DOI: 10.1177/1747954119899880
journals.sagepub.com/home/spo

step counts, energy expenditure and sleep.
13,14
However, only one study in a systematic review
15
reported on the validity of a consumer-wearable in
measuring distance.
16
There is currently a lack of
research on the accuracy of GPS sport watches in the
field of endurance running. With these gaps in mind,
the aim of the current study was to determine the accu-
racy of various GPS sport watches in measuring dis-
tance throughout a hilly 56 km running race.
Methods
Race data
The Two Oceans Marathon is a 56 km road race held in
Cape Town, South Africa. The course includes varia-
tion in terrain including straight open road sections
and a windy mountain pass (Figure 1(a)) and includes
approximately 800 m of cumulative vertical gain
(Figure 1(b)). The 2017 race was held on 15 April.
The weather during the allotted race time ranged
from 14Cto24
C with sunny and clear skies (www.
timeanddate.com). Participants in the study were
recruited from the Two Oceans Marathon list of regis-
tered runners (approximately 10,000 entrants). Upon
registering for the race, runners have the option to
tick ‘yes’ or ‘no’ in terms of being contacted for possi-
ble research studies. Approximately one month before
the race, study personnel sent an e-mail to approxi-
mately 10,000 registered runners who consented ‘yes’
inviting them to participate. All study participants pro-
vided informed consent for their race results and GPS
files to be accessed for the research study. Race results
and segment times of study participants (n ¼255) in
the 2017 race were accessed via the public website
(www.twooceansmarathon.org.za). GPS device files
from participants were accessed via the SmartBeat
Technologies research database (https://www.smart
beatlabs.com), which runners joined by linking
their Strava (www.strava.com) account to the data-
base. The Two Oceans Marathon and SmartBeat
Technologies databases are registered with the
Human Research Ethics Committee at the University
of Cape Town. Segment times were recorded when run-
ners crossed timing mats at the 16 km, 28 km, 42.2 km,
50 km and finish (56 km) points. The segment from the
start to the 16 km mat was excluded from analysis since
runners ran various distances in this segment depend-
ing on their assigned starting batch (batches A–E; with
a time range of 1 min 24 s to 6 min 32 s to cross the first
timing mat).
GPS devices
Runners were divided into eight categories depending
on their GPS device brand (Table 1). Participants pro-
vided their own GPS devices. Eight categories defined
by the GPS device brand are included (Table 1).
Because approximately two-thirds of the participants
had GarminV
Rdevices, they were divided into groups
Figure 1. (a) Two Oceans Marathon 56-km race route. (b) Two Oceans Marathon 56-km race profile. Segment 1 (0–16 km), 2 (16–
28 km), 3 (28–42.2 km), 4 (42.2–50 km), and 5 (50–56 km) are designated with dashed lines.
2International Journal of Sports Science & Coaching 0(0)

according to three unique models. The categories
include Garmin FenixV
Rseries (GFX), Garmin XTV
R
series (GXT), Garmin ForerunnerV
Rseries (GFR),
Activity watches (ACT), SuuntoV
R(STO), TomTomV
R
(TOM), PolarV
R(PLR) and cell phones (CEL). All par-
ticipants in the CEL category used the Strava applica-
tion to record their race. Eight of the 14 CEL
participants used the Strava iPhone app and six used
the Strava android app.
Analysis of GPS device error
Participants were instructed to start their GPS device at
the same time as the starting gun. Their race times were
recorded as they crossed timing mats at the following
race distances: 16 km, 28 km, 42.2 km, 50 km and
56 km. Electronic timing mats and chips were provided
by the software company, RaceTec (http://www.racetec
timing.com/Default.aspx). RaceTec has a Microsoft
SQL server back-end. Timing chips were attached to
the runners’ shoes using their shoelaces. The race route
is measured by an International Association of
Athletics Federation (IAAF) accredited course measur-
er. The times runners crossed the timing mats were used
to identify the corresponding time points in the GPS
file. For example, the Two Oceans race measures a dis-
tance of 12 km in segment 2 of the race (from 16 to
28 km). The Two Oceans database reports runner 
crosses the 16 km mat at 1 h 11 min 17 s and the
28 km mat at 2 h 2 min 26 s. These time points were
identified in the GPS file and the GPS measured dis-
tance between these time points was calculated.
Runners who had no split times available in the race
results or who failed to complete the 56 km race were
excluded from analysis (n ¼26).
A relative error was calculated by taking the course
distance segment (d
course
) subtracted from the GPS
device segment distance (d
GPS
) divided by the average
of the two values (relative error ¼d
GPS
d
course
/
average d
GPS
,d
course
). An error in metres was calculated
by taking the course distance segment subtracted from
the GPS device segment distance (metres error ¼
d
GPS
d
course
). The sign of the error (over or under)
was included in both calculations.
Statistical analysis
All statistical analyses were done using GraphPad
Prism 7 software (GraphPad Software, La Jolla, CA,
USA). Participant race times for each GPS category
are expressed as mean standard deviation (mean 
SD) and were compared using a Kruskal–Wallis test.
Due to the presence of unequal variances, non-
parametric statistical analyses were performed, and all
GPS results are expressed as median interquartile
range (IQR). A Kruskal–Wallis test was used to deter-
mine differences in relative and distance errors between
GPS device categories for segment 2 (16–28 km), seg-
ment 3 (28–42.2 km), segment 4 (42.2–50 km), segment
5 (50–56 km), and the 16–56 km points. Multiple com-
parisons were corrected by controlling the false discov-
ery rate via the two-stage step-up method of
Benjamini et al.
17
Results
The GPS device categories and mean race times for
each category are presented in Table 1. The average
finishing times of runners in the different categories
were not different between groups (p ¼0.20; Kruskal–
Wallis statistic ¼9.76).
The data for segment 1 were not analysed because of
the varying time it took participants to cross the start
line after the start of the race. For segment 2 (16–
28 km), the GXT, GFR, and TOM devices had lower
relative (p ¼0.001; <0.001; <0.001, respectively) and
distance errors (p ¼0.001; p ¼0.001; p <0.001, respec-
tively) compared to the GFX devices (Figure 2(a),
Tables 2 and 3). GXT, GFR, and TOM devices also
had lower relative and distance errors (p <0.001) com-
pared to CEL devices (Figure 2(a), Tables 2 and 3).
For segment 3 (28–42.2 km), all device categories
except for CEL had lower relative (GXT, GFR, STO,
TOM, p <0.001; ACT, p ¼0.024; PLR, p ¼0.005) and
distance errors (GXT, GFR, STO, TOM, p <0.001;
ACT, p ¼0.023; PLR, p ¼0.004) compared to the
GFX (Figure 2(b), Tables 2 and 3). All device catego-
ries except for GFX and ACT had lower relative errors
Table 1. Number of participants and finishing time of GPS Sport
Watch categories.
Device n Race time (h:min:s)
GFX 34 5:52:31 44:39
GXT 58 5:42:50 42:28
GFR 67 5:45:04 50:46
ACT 10 6:12:12 35:13
STO 20 5:42:33 40:46
TOM 44 6:02:58 40:24
PLR 8 5:42:38 52:28
CEL 14 6:04:10 36:42
255
Note: Data are presented as mean SD. GFX series (1, 2, 3); GXT series
(305, 310, 735, 910, 920); GFR series (25, 35, 220, 225, 230, 235, 610,
620, 630); ACT (Garmin VivoActive, Fitbit Surge, Samsung Gear 2 and 3,
Apple Watch); STO (Ambit 2, 2 R, 2S; Ambit 3 Sport, Peak, Run; Spartan);
TOM (Runner 1, 2, 3; Multisport); PLR (M400, V800); CEL (IPhone,
Samsung, Sony). GFX: Garmin FenixV
Rseries; GXT: Garmin XTV
Rseries;
GFR: Garmin ForerunnerV
Rseries; ACT: Activity watches; STO: SuuntoV
R;
TOM: TomTomV
R; PLR: PolarV
R; CEL: cell phones.
Johansson et al. 3

(GXT, GFR, TOM, p <0.001; STO, p ¼0.001; PLR,
p¼0.017) and distance errors (GXT, GFR, TOM,
p<0.001; STO, p ¼0.001; PLR, p ¼0.017) compared
to CEL (Figure 2(b), Tables 2 and 3). TOM had a
lower relative error (p ¼0.030) versus ACT (Figure 2
(b), Table 3).
For segment 4 (42.2–50 km), the GXT, GFR and
TOM devices had higher relative (p <0.001) and distance
errors (p <0.001) versus GFX (Figure 2(c), Tables 2 and
3). All device categories except for PLR and GFX had
significantly different relative (GXT, GFR, TOM,
p<0.001; ACT, p ¼0.005; STO, p ¼0.016) and distance
errors (GXT, GFR, TOM, p <0.001; ACT, p ¼0.004;
STO, p ¼0.013) versus CEL (Figure 2(c), Tables 2 and
3). The STO and PLR devices had lower relative (STO,
p¼0.005; PLR, p ¼0.024) and distance errors (STO,
p¼0.006; PLR, p ¼0.020) versus TOM. The GXT and
GFR watches had higher relative (p ¼0.012, p ¼0.001,
respectively) and distance errors (p ¼0.010, p ¼0.001,
respectively) versus STO. GFR had a higher relative
(p ¼0.011) and distance (p ¼0.010) error versus PLR
(Figure 2(c), Tables 2 and 3).
For segment 5 (50–56 km), GXT, GFR and TOM
devices had lower relative (p ¼0.003; p ¼0.003;
p¼0.002, respectively) and distance errors (p ¼0.002;
p¼0.004; p ¼0.001, respectively) compared to the
CEL devices (Figure 2(d), Tables 2 and 3).
When relative errors were calculated for 16–56 km,
GFX and CEL had higher relative errors compared to
all other devices except for ACT (GFX vs. GXT, GFR,
TOM, respectively, p <0.001; GFX vs. STO, p ¼0.004;
GFX vs. PLR, p ¼0.012) (CEL vs. GXT, GFR, TOM,
respectively, p <0.001; CEL vs. STO, p ¼0.004; CEL
vs. PLR, p ¼0.009) (Figure 3, Table 3). GXT, GFR,
TOM and STO had errors within the IAAF acceptable
course measurement error of 1% (shaded area
(a) (b)
(c) (d)
Figure 2. (a) Relative error of GPS measured distance versus IAAF measured distance for segment 2 (16–28 km). *p <0.05 versus
GFX.
#
p<0.05 versus CEL. (b) Relative error of GPS measured distance versus IAAF measured distance for segment 3 (28–42.2km).
*p <0.05 versus GFX.
#
p<0.05 versus CEL. ^p <0.05 versus TOM. (c) Relative error of GPS measured distance versus IAAF
measured distance for segment 4 (42.2–50 km). *p <0.05 versus GFX.
#
p<0.05 versus CEL. ^p <0.05 versus TOM.
!
p<0.05 versus
STO.
$
p<0.05 versus PLR. (d) Relative error of GPS measured distance versus IAAF measured distance for segment 5 (50–56 km).
#
p<0.05 versus CEL.
4International Journal of Sports Science & Coaching 0(0)

Figure 3). PLR had 6 of 8 participants that fell within
the shaded area; CEL had 2 of 14 participants; and
GFX had 2 of 29. When analysed as distance differ-
ences, GFX and CEL had higher errors compared to
all other devices except for ACT (GFX vs. GXT, GFR,
TOM, respectively, p <0.001; GFX vs. STO, p ¼0.004;
GFX vs. PLR, p ¼0.012) (CEL vs. GXT, GFR, TOM,
respectively, p <0.001; CEL vs. STO, p ¼0.003; CEL
vs. PLR, p ¼0.009) (Table 2).
Discussion
This study has three main findings. Firstly, GPS devi-
ces recorded distance within 0.6 0.3% to 1.9 1.5%
(median IQR) accuracy. Secondly, GXT, GFR, STO,
Table 2. The absolute distance between the GPS sport watch measurement for a given segment(s) versus the IAAF course
measurement.
GFX GXT GFR ACT STO TOM PLR CEL
Segment 2 (m) 80 30 50 30
a,b
50 40
a,b
80 90 60 10 50 20
a,b
50 70 100 100
Segment 3 (m) 640 480 370 120
a,b
390 90
a,b
480 150
a
420 80
a,b
380 60
a,b
420 80
a,b
600 210
Segment 4 (m) 230 90 290 50
a,b,d
300 50
a,b,d,e
250 100 260 40
b,c
290 30
a,b
260 40
c
120 130
Segment 5 (m) 120 40 100 30
b
110 40
b
110 30 120 30 100 20
b
110 40 150 90
Segment 2 –
Segment 5 (m)
640 320 240 210
a,b
240 120
a,b
380 280 290 70
a,b
240 80
a,b
280 160
a,b
760 510
Note: Data are presented as median IQR. Segment 2 ¼16–28 km; segment 3 ¼28–42 km; segment 4 ¼42–50 km; segment 5 ¼50–56 km; segment
2–segment 5 ¼16–finish (40 km). IQR: inter quartile range; GFX: Garmin FenixV
Rseries; GXT: Garmin XTV
Rseries; GFR: Garmin ForerunnerV
Rseries;
ACT: Activity watches; STO: SuuntoV
R; TOM: TomTomV
R; PLR: PolarV
R; CEL: cell phones.
a
Significantly different versus GFX.
b
Significantly different versus CEL.
c
Significantly different versus TOM.
d
Significantly different versus STO.
e
Significantly different versus PLR.
Table 3. The relative difference between the GPS sport watch measurement for a given segment(s) versus the IAAF course
measurement.
GFX GXT GFR ACT STO TOM PLR CEL
Segment 2 (%) 0.70.3 0.5 0.3
a,b
0.4 0.4
a,b
0.7 2.3 0.5 0.1 0.4 0.2
a,b
0.4 0.9 0.8 0.9
Segment 3 (%) 3.03.4 1.2 0.9
a,b
1.3 0.7
a,b
2.0 1.6
a,c
1.5 0.6
a,b
1.3 0.5
a,b
1.5 0.7
a,b
2.8 1.7
Segment 4 (%) 0.4 1.3 1.2 0.7
a,b,d
1.3 0.8
a,b,d,e
0.7 1.8
b
0.8 0.6
b,c
1.2 0.4
a,b,e
0.7 0.5 1.0 2.1
Segment 5 (%) 1.90.7 1.7 0.5
b
1.7 0.7
b
1.9 0.6 1.9 0.5 1.7 0.3
b
1.9 1.0 2.4 1.8
Segment 2–
Segment 5 (%)
1.6 0.9 0.6 0.6
a,b
0.6 0.3
a,b
0.9 1.2 0.7 0.2
a,b
0.6 0.3
a,b
0.7 0.7
a,b
1.9 1.5
Note: Data are presented as median IQR. Segment 2 ¼16–28 km; segment 3 ¼28–42 km; segment 4 ¼42–50 km; segment 5 ¼50–56 km; segment
2–segment 5 ¼16 km–finish (40 km). IQR ¼inter quartile range. GFX: Garmin FenixV
Rseries; GXT: Garmin XTV
Rseries; GFR: Garmin ForerunnerV
R
series; ACT: Activity watches; STO: SuuntoV
R; TOM: TomTomV
R; PLR: PolarV
R; CEL: cell phones.
a
Significantly different versus GFX.
b
Significantly different versus CEL.
c
Significantly different versus TOM.
d
Significantly different versus STO.
e
Significantly different versus PLR.
Figure 3. Relative error of GPS measured distance versus IAAF
measured distance for segment 2–segment 5 (16–56 km).
Note: Data are presented as median +/- IQR.
Johansson et al. 5

TOM and PLR have lower overall errors compared to
CEL and GFX. Thirdly, the errors appear to be greater
in the hilly sections of the race with more turns (seg-
ments 3, 4 and 5) compared to the flatter and straighter
section (segment 2).
When considered in context, the relative error for
segment 2 to the finish of the race (i.e. a distance of
40 km) was lower than previous research using a GPS
sport watch.
18
Nielsen et al.
18
showed the error in mea-
suring distance using a Garmin Forerunner 110 ranged
from 0.8% to 6.2% (mean). The results from our study
reflect better accuracy. This could be due to advance-
ments in technology and GPS units since 2013 when the
Nielsen paper was published.
The cell phone category displayed a higher error
compared to most other categories. Eight participants
had iPhones, and six participants had phones with
android operating systems. Different brands may
have unique GPS units. However, when iPhones were
compared to cell phones with android operating sys-
tems, there was no difference in relative error for the
16–56 km segment (p ¼0.57). There has also been
research about the accuracy of different running appli-
cations on cell phones.
19
All the study participants in
the CEL category used the Strava running application
(www.strava.com); therefore, this possibility was elim-
inated. The placement of the GPS device on the body
can affect accuracy. GPS sport watches are usually
worn on the wrist with a clear view of the sky. In con-
trast, cell phones can be held or worn on the arm with a
case and strap, in a shorts pocket or in a waistband.
The placement of the cell phone may have contributed
to the large IQR of the CEL category.
The Garmin Fenix category also displayed a higher
error compared to most other categories. The third edi-
tion of the Garmin Fenix series (31 of 34 participants in
this category had the Garmin Fenix 3) is the first one
to offer GPS and Global Navigation Satellite
Systems (GLONASS).
20
The recent incorporation of
GLONASS into GPS sport watches increases the avail-
ability of surrounding satellites. However, there is cur-
rently no research comparing the accuracy of GPS
versus GPS combined with GLONASS in the devices
included in our study. Simply increasing the availability
of surrounding satellites does not guarantee improved
GPS accuracy. There are many factors to consider such
as the quality of the GPS chipset in the device as well as
quality of the manufacturer algorithm. Future research
should investigate these areas further.
The next finding was that the error appears to
change depending on the segment of the race course.
For example, the medians in the TomTom category for
segments 2, 3, 4 and 5 are 0.4%, 1.3%, 1.2% and
1.7%, respectively. This change in error could be due
to changes in course elevation, turns and/or tree cover.
Segment 2 is relatively straight and flat with 100 m of
elevation change while segments 3, 4 and 5 include
larger changes in elevation (465 m, 300 m and
150 m, respectively) and more turns. Nielsen et al.
18
tested the accuracy of the Garmin Forerunner 110 in
reading distance covered in three different settings. The
relative error on a flat path with a clear view of the sky
was 0.8%. In an urban area with buildings, the error
was 1.2%, and in a covered forest area, the error was
6.2%. Segments 4 and 5 include significant tree cover
compared to the other segments which may have affect-
ed the accuracy. In addition, segments 3 and 4 include
stretches of road surrounded by mountains on one or
both sides which can also affect accuracy. A profes-
sional product tester found an average GPS sport
watch error of 1% on a 1.6 km straight path with a
clear view of the sky; 1.7% around a 400 m track and
1.7% on a 0.84 km loop route through a mix of open
area and tree cover.
21
In line with this, there is a pos-
sibility the turns in segments 3, 4 and 5 (Figure 1(a))
affected the accuracy in reading distance.
Limitations
A limitation of the study is that the distance between
course markings can only be guaranteed to 1%
according to the IAAF regulations. This limitation
was considered in the analysis and interpretation.
Consideration should be given to the various software
and firmware within devices as both can affect GPS
accuracy. Because we do not know which software ver-
sion participants had on their devices on race day, it is
not possible for us to speculate about this. Regardless
of whether we categorized by brand, hardware or soft-
ware, there would have been cross-over between hard-
ware and software. The main aim of our study was to
determine if using GPS sport watches are feasible and
valid method to measure distance. We feel reporting
results by brand has more practical applications for
sport scientists, coaches and athletes as firmware is
not included in the marketed specifications of the devi-
ces. Another limitation of the study includes the lack of
control of device settings such as amount of free
memory and sampling rate which can affect accuracy.
The possibility of runners choosing a low sampling rate
is low as the battery life using GPS for 29 of 38 devices
is 10 h or more. With a race cut-off time of 7 h, having
to alter the sampling rate to extend battery life would
not be a likely occurrence in this race. Finally, the lack
of control of the placement of the GPS device is a lim-
itation of this study, particularly in the CEL category.
GPS sport watches are predominantly worn on the
wrist with a clear view of the sky. Cell phones are
often worn around the arm in a case or at the waist
in a pocket both of which can affect accuracy. While
6International Journal of Sports Science & Coaching 0(0)

this limitation needs to be considered when interpreting
the results, it has practical implications as it is not fea-
sible for runners to carry cell phones without a case
and/or strapping it to an arm or placing in a pocket.
Conclusion
The conclusions of the current study are that (1) GPS
sport watches are a valid and feasible way to measure
performance and track training load; (2) GXT, GFR,
STO, TOM and PLR devices have lower overall errors
compared to CEL and GFX devices; and (3) the error
appears to be greater in segments with more elevation
and turns (segments 3, 4 and 5 of the race) compared to
the flatter and straighter section (segment 2). The GPS
sport watches in this study have an accuracy of
0.6 0.3% to 1.9 1.5% (median IQR) in reporting
distance covered. This indicates that GPS sport
watches are a valid and feasible method for sport sci-
entists and coaches to measure performance and track
training load. However, the small error associated with
each brand needs to be considered when data are inter-
preted. Sport scientists, coaches and runners should
also consider that error may be influenced by changes
in elevation, non-linear movement, tree cover and
cloud cover.
Acknowledgements
The authors sincerely thank all the research participants. The
authors also thank the Two Oceans Marathon Race
Committee and SmartBeat Technologies.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with
respect to the research, authorship, and/or publication of this
article.
Funding
The author(s) disclosed receipt of the following financial sup-
port for the research, authorship, and/or publication of this
article: MIL acknowledges financial support from the National
Research Foundation incentive fund. REJ acknowledges
financial support from the University of Cape Town
International Student Scholarship Fund.
ORCID iD
Rebecca E Johansson https://orcid.org/0000-0002-0605-
1293
References
1. Duking P, Hotho A, Holmberg HC, et al. Comparison of
non-invasive individual monitoring of the training and
health of athletes with commercially available wearable
technologies. Front Physiol 2016; 7: 71.
2. Koytcheva M. Wearables market to be worth $25 billion
by 2019. CCS Insight, www.ccsinsight.com/press/compa
ny-news/2332-wearables-market-to-be-worth-25-billion-
by-2019-reveals-ccs-insight (2015, accessed 31 October
2017).
3. Cummins C, Orr R, O’Connor H, et al. Global position-
ing systems (GPS) and microtechnology sensors in team
sports: a systematic review. Sports Med 2013; 43:
1025–1042.
4. Schutz Y and Chambaz A. Could a satellite-based
navigation system (GPS) be used to assess the physical
activity of individuals on earth? Eur J Clin Nutr 1997; 51:
338–339.
5. Baranski P and Strumillo P. Enhancing position accuracy
in urban terrain by fusing data from GPS receiver, iner-
tial sensors, stereo-camera and digital maps for pedestri-
an navigation. Sensors 2012; 12: 6764–6801.
6. Akenhead R, French D, Thompson KG, et al. The accel-
eration dependent validity and reliability of 10Hz GPS.
J Sci Med Sport 2014; 17: 562–566.
7. Aughey RJ. Applications of GPS technologies to field
sports. Int J Sports Physiol Perform 2011; 6: 295–310.
8. Barbero-Alvarez JC, Coutts A, Granda J, et al. The
validity and reliability of a global positioning satellite
system device to assess speed and repeated sprint ability
(RSA) in athletes. J Sci Med Sport 2010; 13: 232–235.
9. Cardinale M and Varley MC. Wearable training
monitoring technology: applications, challenges, and
opportunities. Int J Sports Physiol Perform 2017; 12:
S2-55–S2-62.
10. Chambers R, Gabbett TJ, Cole MH, et al. The use of
wearable microsensors to quantify sport-specific move-
ments. Sports Med 2015; 45: 1065–1081.
11. Malone JJ, Lovell R, Varley MC, et al. Unpacking the
black box: applications and considerations for using GPS
devices in sport. Int J Sports Physiol Perform 2017; 12:
S2-18–S2-26.
12. Abt G, Bray J and Benson AC. The validity and inter-
device variability of the Apple Watch
TM
for measuring
maximal heart rate. J Sports Sci 2017; 3: 1–6.
13. Diaz KM, Krupka DJ, Chang MJ, et al. Fitbit: an accu-
rate and reliable device for wireless physical activity
tracking. Int J Cardiol 2015; 185: 138–140.
14. Ferguson T, Rowlands AV, Olds T, et al. The validity of
consumer-level, activity monitors in healthy adults worn
in free-living conditions: a cross-sectional study. Int J
Behav Nutr Phys Act 2015; 12: 42.
15. Evenson KR, Goto MM and Furberg RD. Systematic
review of the validity and reliability of consumer-
wearable activity trackers. Int J Behav Nutr Phys Act
2015; 12: 159.
16. Takacs J, Pollock CL, Guenther JR, et al. Validation of
the Fitbit One activity monitor device during treadmill
walking. J Sci Med Sport 2014; 17: 496–500.
17. Benjamini Y, Krieger AM and Yekutieli D. Adaptive
linear step-up procedures that control the false discovery
rate. Biometrika 2006; 93: 491–507.
18. Nielsen RO, Cederholm P, Buist I, et al. Can GPS be
used to detect deleterious progression in training
Johansson et al. 7

volume among runners? J Strength Cond Res 2013; 27:
1471–1478.
19. Bauer C. On the (in-) accuracy of GPS measures of
smartphones: a study of running tracking applications.
In: Proceedings of the 11th International Conference on
Advances in Mobile Computing & Multimedia
(MoMM2013), Vienna, Austria, 2–4 December 2013,
pp. 335–340. DOI:10.1145/2536853.2536893.
20. Hofmann-Wellenhof B, Lichtenegger H and Wasle E.
GNSS–global navigation satellite systems: GPS,
GLONASS, Galileo, and more. Morlenbach: Springer-
Verlag Wien, 2007.
21. Rainmaker DC. 2011 Sport Device GPS Accuracy in
Depth: Part I [internet], www.dcrainmaker.com/2011/
06/2011-sport-device-gps-accuracy-in-depth.html (2011,
accessed 4 October 2019).
8International Journal of Sports Science & Coaching 0(0)