Validity and reliability of a motorized sprint resistance device

Abstract

An increasing number of sprint-related studies have employed robotic devices to provide resistance while sprinting. The aim of this study was to establish within-session reliability and criterion validity of sprint times obtained from a robotic resistance device. Seventeen elite female handball players (22.9 ± 3.0 y; 176.5 ± 6.5 cm; 72.7 ± 5.5 kg; training volume 9.3 ± 0.7 hrs per week) performed two 30-m sprints under three different resistance loading conditions (50, 80 and 110 N). Sprint times (t0-5m, t5-10m, t10-15m, t15-20m, t20-30m and t0-30m) were assessed simultaneously by a 1080 Sprint robotic resistance device and a post-processing timing system. The results showed that 1080 Sprint timing was equivalent to the post-processing timing system within the limits of precision (± 0.01 s). A systematic bias of ~ 0.34 ± 0.01 s was observed for t0-5m caused by different athlete location and velocity at triggering point between the systems. Coefficient of variation was ~ 2% for t0-5 and ~ 1% for the other time intervals, while standard error of measurement ranged from 0.01 to 0.05 s, depending on distance and phase of sprint. Intraclass correlation ranged from 0.86 to 0.95. In conclusion, the present study shows that the 1080 Sprint is valid and reliable for sprint performance monitoring purposes.

Full text

Sprint-time measurements 1
Validity and reliability of a robotic sprint resistance device
Elvir Rakovic1, Gøran Paulsen2,3, Christian Helland2, Thomas Haugen4, Ola Eriksrud3
1 Department of Food and Nutrition and Sport Science, University of Gothenburg, Gothenburg, Sweden.
2 Norwegian Olympic and Paralympic Committee and Confederation of Sports, Oslo, Norway.
3 Department of Physical Performance, Norwegian School of Sports of Science, Oslo, Norway.
4 School of Health Sciences, Kristiania University College
Abstract
An increasing number of sprint-related studies have employed robotic devices to provide
resistance while sprinting. The aim of this study was to establish within-session reliability and
criterion validity of sprint times obtained from a robotic resistance device. Seventeen elite
female handball players (22.9  3.0 y; 176.5  6.5 cm; 72.7  5.5 kg; training volume 9.3  0.7
hrs per week) performed two 30-m sprints under three different resistance loading conditions
(50, 80 and 110 N). Sprint times (t0-5m, t5-10m, t10-15m, t15-20m, t20-30m and t0-30m) were assessed
simultaneously by a 1080 Sprint robotic resistance device and a post-processing timing system.
The results showed that 1080 Sprint timing was equivalent to the post-processing timing system
within the limits of precision ( 0.01 s). A systematic bias of ~ 0.34 ± 0.01 s was observed for
t0-5m caused by different athlete location and velocity at triggering point between the systems.
Coefficient of variation was ~ 2% for t0-5 and ~ 1% for the other time intervals, while standard
error of measurement ranged from 0.01 to 0.05 s, depending on distance and phase of sprint.
Intraclass correlation ranged from 0.86 to 0.95. In conclusion, the present study shows that the
1080 Sprint is valid and reliable for sprint performance monitoring purposes.
Key words: Spatiotemporal measurements; sprint conditioning; photocells; resisted sprinting.

Sprint-time measurements 2
INTRODUCTION
Sprint training and testing are common routines for many athletes and coaches. Such practices
are accompanied by a variety of modalities (e.g., linear or change-of-direction sprints,
accelerated or maximal velocity sprinting), loading components (duration, intensity, resting
periods, session rate, resisted/assisted conditions, etc.), procedures (e.g., time initiation and
starting position) and equipment (timing gates, laser guns and radar devices, GPS, sleds, towing
cords, footwear, etc.) (3, 4).
An increasing number of sprint-related studies have employed robotic devices to provide
resistance while sprinting, with the 1080 Sprint (1080 Motion AB, Stockholm, Sweden)
commonly used (1, 7-9, 11). Application of such a device may serve several benefits. Firstly,
an accurate resistance can be predetermined, which is more challenging with e.g. sleds due to
surface friction issues under varying environmental conditions. Moreover, synchronized
assessments of velocity and displacement relative to the start line with the force exerted through
the machine’s cord under varying loading conditions can be obtained by one device only. This
will negate the need for the combination of sleds with photocells, laser guns or radars. The
distance-time or velocity-time running data can in turn be used for computation of macroscopic
mechanical outputs (10) that may form basis for individual training prescription (1, 9, 10).
However, these potential benefits are dependent on the ability of the robotic device to
accurately assess velocity-time data. To the best of our knowledge, no studies to date have
addressed this issue. The purpose of this study was therefore to determine within-session
reliability and criterion validity of sprint split times obtained from a 1080 Sprint robotic device.
METHODS
Experimental approach to the problem
The data used for this reliability and validation study were compiled from anonymized data
from a previously published investigation exploring the effect of individual sprint training
prescription based on force-velocity (FV) profiles (9). Because it is crucial that the entire
acceleration phase of sprinting athletes is covered by timing gates to ensure valid and reliable
FV profiles (10), the female elite team sport athletes performed 30-m sprints with varying
resistance loading. Split times (t0-5m, t5-10m, t10-15m, t15-20m, t20-30m and t0-30m) were assessed
simultaneously by a robotic resistance device and a post-processing timing system. These
measurements formed basis for intra-session reliability and validity assessments.

Sprint-time measurements 3
Subjects
Seventeen elite female handball players (mean  SD: 22.9  3.0 years; 176.5  6.5 cm; 72.7 
5.5 kg; total training volume 9.3  0.7 hrs per week) with a minimum of 10-y handball-specific
conditioning volunteered to participate. Four of these played for the national team while eleven
players participated in the Champions League tournament during the current season. The study
was reviewed by the Regional Ethics Committee and approved by the Norwegian Data
Protection Authority. Due to the newly implemented General Data Protection Regulations
(GDPR) by the European Union, the local university XXXX XXXX XXXX XXXX has the
responsibility for data security and ethics. All participants signed an informed consent form
prior to participation, and this study was conducted according to the Declaration of Helsinki.
Procedures
A standardized 20-min warm-up consisting of jogging (~60–75% of age-predicted maximal
heart rate), selected exercises (lunges, hip lift, ballistic mobility hamstrings and hips in prone
and supine), running drills (high knees, skipping, butt-kicks, straight leg pulls) and three to four
sprints with increasing speed was conducted prior to testing (9). After the warm-up, the athletes
performed two maximal 30-m sprints with 50, 80 and 110 N resistance respectively, in a
randomized order (i.e., one sprint with each resistance before proceeding to the next sequence).
The resistance during the six sprints was provided by a 1080 Sprint robotic device (1080
Motion AB, Stockholm, Sweden). All sprints were initiated from a standing, split-stance
position with the tip of the toe of the front foot placed on the start line. All starts were
commenced from a static position, meaning that “leaning backward before rolling forward”
was not allowed. After a ready signal was given by the test leader, the athletes started on their
own initiative. Recovery time between each sprint was ~ 4 min.
MuscleLab timing system (Ergotest AS, Porsgrunn, Norway) was used to assess sprint times.
An infrared optical contact mat covered the start line, and timing was initiated at the point of
front foot lift-off. Post-processing timing gates (i.e., an internal software scans all signals from
the timing gate in terms of frequency and duration) where mounted on tripods 120 cm above
floor level and placed at 5, 10, 15, 20 and 30 m. Thus, all timing gates were mounted above
hip height to avoid undue beam break caused by the lower limbs (3). The onset of the longest
break of the infrared beam was used as a trigger criterion, as the torso will produce a longer
break than an arm (3). Earp & Newton reported that the signal processing technology
completely removed all false signals (i.e., time triggering caused by swinging limbs) (2).

Sprint-time measurements 4
Moreover, Rakovic et al. reported excellent reliability values for this system setup, as typical
error (TE) and coefficient of variation (CV) were 0.03 s and 1.0% for 0–30 m sprint time and
0.08 m∙s−1 and 1.4% for V0 (9). Hence, the MuscleLab timing system was used as gold standard
for sprint performance assessments in this study.
The 1080 Sprint was used to provide resistance and assess sprint times. This portable system
uses a servo motor (2000 RPM OMRON G5 Series Motor, OMRON Corporation, Kyoto,
Japan) to provide resistance while sprinting. The robotic device was placed 5 m behind the
starting line with the line attached to the athlete by a centrally located ring (sacrum) on a belt
firmly tightened around the pelvis. The resistance load (50, 80 or 110 N) was determined and
controlled by the computer application (1080 Motion, Lidingö, Sweden). The isotonic
resistance mode was used, as different modes are offered by the 1080 Sprint. Position trigger
criterion for time initiation was set to 30 cm of line being pulled away from the machine. This
corresponds to the position of the pelvis being ~ 30 cm past the start line. Data (force, position
and time) were recorded at 333 Hz.
Statistical analysis
Mean and standard deviation are presented for all sprint times. Shapiro-Wilk test was used to
test the assumption of normality for each set of sprint time data, and z-scores were calculated
and analyzed for both skewness and kurtosis. Intraclass correlation coefficient (ICC), standard
error of measurement (SEM) and coefficient of variation (CV) were calculated for all sprint-
time intervals to determine within session reliability. Criterion validity was based on mean
difference (tdiff), CV and Pearson’s r correlation. Spearman’s rank correlation was used instead
of Pearson’s r where the datasets were not normally distributed. Bland Altman plots were
created for sprint-time difference distribution between the timing systems.
RESULTS
****Table 1 about here****
****Figure 1 about here****
Table 1 shows within session reliability and criterion validity for 1080 Sprint. Regarding
reliability, CV ranged from 1.93 to 2.56% for t0-5 and from 0.82 to 1.34 for the other time
intervals, while SEM ranged from 0.01 to 0.05, depending on distance and phase of sprint. ICC
ranged from 0.86 to 0.95.

Sprint-time measurements 5
Distribution of sprint time differences for all resisted sprints are presented in Figure 1. Biases
(tdiff ) where low for t5-10m, t10-15m, t15-20m and t20-30m (range = -0.01 to 0.01 s) for all resistance
conditions. Greater differences were observed for t0-5m (range = 0.33 to 0.35 s) and t0-30m (range
= 0.31 to 0.34 s) across all resistance conditions.
DISCUSSION
The aim of the present study was to explore within session reliability and criterion validity of
sprint split times obtained from a robotic device during resisted sprinting. Overall, the 1080
Sprint device displayed satisfactory reliability values. The reliability values observed are
comparable to previously validated and commonly used timing systems (3). The poorest values
were observed for t0-5m. This is in line with Haugen & Buchheit (3), who reported considerably
poorer reliability (typical error) for t0-5m compared to longer sprint-distance intervals.
The current analysis revealed no systematic variation between the 1080 Sprint and the post-
processing timing gates, except for t0-5m and t0-30m. That is, for practical purposes these systems
give similar results to a precision of ± 0.01 s. Post-processing timing gates, which were used
as gold-standard in this case, are considered accurate for sprint performance monitoring, as the
internal software processes and remove false signals (3). This provides that the timing gates
are mounted above hip height (to avoid undue beam break caused by the lower limbs), as
performed in this study. However, a systematic bias of ~ 0.34 ± 0.01 s was observed for t0-5m
and t0-30m. This is not surprising, as the starting method and timing system used can combine
to generate large absolute differences in “sprint time” (3, 6). The sources of time differences
usually include the starting device, vertical and horizontal placement of starting device relative
to the start line, body configuration and velocity at triggering point (6). In this case, pelvis was
~ 60 cm past the start line at time initiation for the optical contact mat (front foot lift-off), while
only ~ 30 cm past the start line at time initiation for the robotic device. Hence, pelvis was ~ 30
cm further past the start line at time initiation for the optical contact mat than for the robotic
device. Provided that the bias is systematic so that correction factors can be generated (as in
this case), sprint performance comparisons across systems can be performed (3). The same
issue is present for calculation of sprint mechanical outputs based on distance-time or speed-
time data. An essential point when using the simple method proposed by Samozino et al. (10)
is that the time 0 must be very close to the first rise of the force production onto the ground.
This is equivalent to a setup with starts from blocks and audio signal with reaction time
subtracted from the total time (5). According to Haugen & Buchheit (3), front-foot triggering

Sprint-time measurements 6
generates 0.51 s faster sprint times compared to starts from blocks where reaction time is
subtracted from the total time. Because the current systematic bias was 0.34 s on average (Table
1), we estimate that a correction factor of ~ 0.17 s (i.e., 0.51 minus 0.34 s) should be added to
the 1080 Sprint times to ensure valid computations of sprint mechanical outputs.
PRACTICAL APPLICATIONS
The present study shows that the 1080 Sprint is valid and reliable for sprint performance
monitoring purposes. This means that multiple functions for sprint training, testing and
monitoring can be operated by one device only. The benefits of using one system in both
research and field based settings includes i) accurate prescription of resistance while obtaining
synchronized assessments of velocity, acceleration and pulling force as a function of time or
displacement relative to starting line, ii) the possibility to apply varying resistance loading
during specific portions of the sprint, iii) monitor individual and team responses (i.e fatigue)
and iiii) computation of sprint mechanical outputs.
AKNOWLEDGEMENTS
Ola Eriksrud is a shareholder in 1080 Motion AB.

Sprint-time measurements 7
REFERENCES
1. Cross MR, Lahti J, Brown SR, Chedati M, Jimenez-Reyes P, Samozino P, Eriksrud O,
and Morin JB. Training at maximal power in resisted sprinting: Optimal load determination
methodology and pilot results in team sport athletes. PloS one 13, 2018.
2. Earp JE and Newton RU. Advances in electronic timing systems: considerations for
selecting an appropriate timing system. J Strength Cond Res 26: 1245-1248, 2012.
3. Haugen T and Buchheit M. Sprint Running Performance Monitoring: Methodological
and Practical Considerations. Sports Med 46: 641-656, 2016.
4. Haugen T, Seiler S, Sandbakk O, and Tonnessen E. The Training and Development of
Elite Sprint Performance: an Integration of Scientific and Best Practice Literature. Sports Med
Open 5: 44, 2019.
5. Haugen TA, Breitschadel F, and Samozino P. Power-Force-Velocity Profiling of
Sprinting Athletes: Methodological and Practical Considerations When Using Timing Gates. J
Strength Cond Res, 2018.
6. Haugen TA, Tonnessen E, and Seiler SK. The difference is in the start: impact of timing
and start procedure on sprint running performance. J Strength Cond Res 26: 473-479, 2012.
7. Helland C, Haugen T, Rakovic E, Eriksrud O, Seynnes O, Mero AA, and Paulsen G.
Force-velocity profiling of sprinting athletes: single-run vs. multiple-run methods. Eur J Appl
Physiol 119: 465-473, 2019.
8. Mangine GT, Huet K, Williamson C, Bechke E, Serafini P, Bender D, Hudy J, and
Townsend J. A Resisted Sprint Improves Rate of Force Development During a 20-m Sprint in
Athletes. J Strength Cond Res 32: 1531-1537, 2018.
9. Rakovic E, Paulsen G, Helland C, Eriksrud O, and Haugen T. The effect of
individualised sprint training in elite female team sport athletes: A pilot study. J Sports Sci 36:
2802-2808, 2018.
10. Samozino P, Rabita G, Dorel S, Slawinski J, Peyrot N, Saez de Villarreal E, and Morin
JB. A simple method for measuring power, force, velocity properties, and mechanical
effectiveness in sprint running. Scand J Med Sci Sports, 2015.
11. Thompson KMA, Whinton AK, Ferth S, Spriet LL, and Burr JF. Moderate Load
Resisted Sprints do Not Improve Subsequent Sprint Performance in Varsity Level Sprinters. J
Strength Cond Res, 2018.

Sprint-time measurements 8
Figure 1. Bland Altman analysis of sprint times (34 trials for each resisted condition) derived from timing gates and 1080 Sprint.

Sprint-time measurements 9
Table 1. Within session reliability and criterion validity for 1080 Sprint
Sprint times (mean ± SD)
Reliability
Criterion validity
Resistance
Interval
tML (s)
t1080 (s)
CV
(%)
SEM
(s)
ICC
tdiff (s)
CV (%)
Cor.
50 N
t0-5m
0.97±0.04
1.30 ± 0.05
2.26
0.03
0.81
.329
20.54
0.79
t5-10m
0.85±0.03
0.85 ± 0.02
1.01
0.01
0.92
.002
1.65
0.75
t10-15m
0.78±0.02
0.77 ± 0.02
1.06
0.01
0.90
-.008
2.01
0.48
t15-20m
0.72±0.02
0.73 ± 0.02
1.03
0.01
0.92
.007
1.60
0.73
t20-30m
1.43±0.04
1.43 ± 0.04
0.82
0.01
0.95
.001
0.74
0.94
t0-30m
4.75±0.11
5.06 ± 0.12
0.91
0.05
0.93
.308
4.49
0.94
80 N
t0-5m
1.03±0.04
1.36 ± 0.05
1.93
0.02
0.87
.331
19.76
0.57
t5-10m
0.90±0.03
0.91 ± 0.02
0.99
0.01
0.92
.005
1.37
0.80
t10-15m
0.83±0.02
0.82 ± 0.02
1.13
0.01
0.89
-.007
1.56
0.71
t15-20m
0.78±0.02
0.79 ± 0.02
1.34
0.01
0.85
.010
1.50
0.79
t20-30m
1.55±0.05
1.56 ± 0.04
1.32
0.02
0.89
.005
0.76
0.93
t0-30m
5.08±0.13
5.41 ± 0.14
1.12
0.04
0.90
.320
5.97
0.94
110 N
t0-5m
1.06±0.05
1.41 ± 0.07
2.56
0.03
0.86
.353
20.49
0.74
t5-10m
0.96±0.04
0.96 ± 0.02
1.10
0.01
0.90
-.002
1.52
0.82
t10-15m
0.89±0.02
0.88 ± 0.02
1.01
0.01
0.92
-.006
1.54
0.68
t15-20m
0.85±0.03
0.86 ± 0.02
0.98
0.01
0.94
.012
1.86
0.83
t20-30m
1.70±0.05
1.71 ± 0.05
1.30
0.02
0.91
.004
0.87
0.92
t0-30m
5.45±0.15
5.79 ± 0.16
1.10
0.04
0.92
.338
4.29
0.95
tML = sprint times from the MuscleLab timing system, t1080 = sprint times from the 1080 Sprint robotic device, CV = coefficient of variation, SEM
= standard error of measurement, ICC = intraclass correlation coefficient, tdiff = time difference between the analyzed systems, Cor. = Correlation
(Pearsons r or Spearman’s rank).