The accuracy of measuring velocity during weightlifting movements with five velocity-based training devices

Abstract

The use of weightlifting exercises is prevalent in competitive and recreational environments, as well as sport-specific training. Traditionally, weightlifting coaches prescribe specific training loads based on an individual's maximal ability. Velocity-based training offers an alternative method that promises to quantify strength based on velocity and provides information that increases competitiveness through real-time feedback. Various velocity measurement devices are available on the market. Their precision is critical for the adequate implementation of velocity-based training. The aim of the present study was to compare the concentric peak velocity measurements of five of these devices during two weightlifting movements, the snatch and clean, to data collected with a 12-camera motion capture system, which was considered as gold standard. It was hypothesized that the velocity measurement devices used in this study would vary in accuracy based on their retail prices. Velocity readings associated with light and moderate (40% and 70% of one-repetition max) loads were measured for both the snatch and clean performed by 12 competitive weightlifters. A least products regression was used to assess validity by comparing five devices against a criterion measure. A general linear model showed statistical differences in the velocities measured with these five devices (p < 0.001). Specifically, the GymAware RS linear position transducer was the most accurate device, demonstrating no fixed or proportional bias when used to quantify velocity during the snatch and clean. The remaining four devices significantly underestimated peak velocity, which would directly impact the daily planning of lifters’ training. Practitioners must consider the error and bias of each device before implementing velocity-based training.

Keywords

Barbell clean linear position transducer motion analysis resistance training snatch

Introduction

Weightlifting¹ is an effective method of enhancing strength, power, and lean body mass,^2–6 and therefore, it is widely prevalent in competitive and recreational resistance training, as well as sport-specific training. In the sport of weightlifting, athletes participate in contests featuring the snatch and clean and jerk, both ground-to-overhead lifts (refer to Figure 1). Success in these lifts, especially as the load increases, hinges on the athlete's ability to harness explosive strength and exhibit rapid body movements.⁷

Figure 1.

Concentric (“pull”) phases of the Olympic-style lifts: (A) snatch and (B) clean and jerk.

Figure 2.

Velocity-based training (VBT) devices: (A) GymAware RS, (B) Vitruve, (C) GravityBox, (D) GymAware FLEX, and (E) WL Analysis.

Traditionally training programs for weightlifting are often formulated based on percentages of an individual's one-repetition maximum (1RM) load for a given lift.^4,5 The 1RM represents the maximum load that the individual can lift for one repetition through a previously determined range of motion,^8,9 and it is often used for assessing maximal strength and gauging intensity.^8,10,11 Indeed, some training systems are predicated on using daily 1RMs to guide the training process.^7,12,13 An alternative programming approach that has recently gained attention within the scientific community is the use of velocity-based training (VBT) methods.^4,14–17 It has been proposed that this method may allow the coach to estimate the 1RM load on a day-to-day basis based on changes in velocity, so that training outcomes and efficiency are maximized.^4,8,10 Specifically, peak velocity and mean velocity during the concentric phase are used to provide the most valuable information when implementing VBT during resistance training.^4,18 Additionally, velocity monitoring allows for real-time feedback during a training session,¹⁹ promoting competition and motivation for athletes.²⁰ Notably, it is important to mention that results from several research studies do not support VBT as a method for accurately estimating 1RM and prescribing daily training in weightlifting.^8,21–24 Moreover, there is an emerging body of evidence that the 1RM does not vary dramatically from day-to-day,^25–27 and that there is greater day-to-day variability in the velocity of movement even though the 1RM and xRM (i.e. 3RM and 6RM) are relatively stable across a microcycle.^24,28,29 Hence, the intention of the present contribution is not to advocate VBT as a superior training methodology, but to strictly assess the accuracy of those tools that are used to measure the bar velocity.

To successfully implement VBT, it is critical that velocity monitoring devices provide users with accurate values.^30,31 One of the most common and accurate devices used to calculate velocity during resistance training is the linear position transducer (LPT), which directly measures the vertical displacement of a cable attached to the barbell.⁶ LPTs vary in price, ranging from $180 to almost $2000. Prior research has examined the validity of LPTs. The GymAware RS (GymAware, Braddon, Australia) LPT (GYM)³² has undergone validation in several studies measuring peak velocity during a variety of resistance exercises.^33–37 Furthermore, the Vitruve (Vitruve, Madrid, Spain) LPT (VIT)³⁸ has been reported to provide valid peak velocities during the back squat.³⁹ Another available LPT in the marker is the GravityBox (Gravity Box, Kelowna, Canada) (GRB).⁴⁰ However, unlike GYM and VIT, to date there are no known studies that have validated the GRB.

Another device by the company GymAware, Kinetic FLEX (GymAware, Braddon, Australia) (FLX), uses a laser optic array to wirelessly track barbell velocity at 50 Hz.⁴¹ When compared to the GYM, the FLX provides a cheaper alternative ($495) for measuring velocities of movement during resistance training.⁴¹ However, when examining the available literature reporting on the validity of the FLX, there is conflicting data on the accuracy of this device.^33,42 In addition, most of these devices require separate app fees and/or subscriptions for premium analytical capabilities of the data collected.^32,38,41 Finally, standalone mobile phone apps which typically cost tens of dollars are a relatively new means for VBT in which a phone-recorded video of the lift is analyzed by in-app software.⁴³ For example, the premium version of WL Analysis (WL Analysis, Ropczyce, Poland) (WLA) is a one-time fee of $9 but has yet to be scientifically validated.⁴⁴ Older apps such as PowerLift (PowerLift, San Francisco, USA) and MyLift (TrainerFu, Fermont, USA) have shown conflicting results for determining velocity for powerlifting exercises.^45–47 These apps offer a better economic value for the implementation of VBT as they can have a sample rate of up to 60 frames per second depending on the mobile device available. However, they lack extensive validation, especially in the sport of weightlifting. Details about each device's sampling rate, cost, and type are reported in Table 1.

Table 1.

Information and features of VBT devices.

Device	Abbreviation	Company	Type	Sampling rate	Cost (US$)
GymAware RS	GYM	GymAware, Braddon, Australia	LPT	100 Hz	1995.00 ³²
Vitruve	VIT	Vitruve, Madrid, Spain	LPT	100 Hz	447.00 ³⁸
GravityBox	GRB	Gravity Box, Kelowna, Canada	LPT	100 Hz	183.22 ⁴⁰
Kinetic FLEX	FLX	GymAware, Braddon, Australia	Laser	50 Hz	495.00 ⁴¹
WL Analysis	WLA	WL Analysis, Ropczyce, Poland	App	30 fps*	9.00 ⁴⁴

VBT: velocity-based training; LPT: linear position transducer.

*WL Analysis videos were recorded at the highest video quality processed by the app at the time of data collection (30 fps).

The aim of this study was to assess the accuracy in measuring peak velocity during weightlifting movements, specifically the snatch and clean. This evaluation was conducted using five VBT devices while lifting loads at 40% and 70% of the 1RM. The assessment involved comparing mean peak velocities recorded with GYM, VIT, GRB, and FLX devices, as well as the WLA mobile phone app, in relation to a motion capture system. Given the diversity in the methods for calculating velocity among the various commercially available devices, variations in the accuracy of the devices tested were anticipated. It was first hypothesized that the GYM would display a degree of validity similar to the motion capture system since it has the highest retail price and has previous scientific validation. We also expect that the mobile app WLA would not display a degree of validity significantly differently compared to the other devices tested regardless of being the most economical option. This is due to the comparable sample rate capability with respect to the LPT and laser devices. This potential finding could benefit practitioners, providing them with a cost-effective option for monitoring barbell velocity and implementing VBT methods. Moreover, it was also hypothesized that these velocity monitoring devices, except GYM, would not be as accurate as the motion capture system, which records with 12 high-speed (80 frames/s) cameras.

Methods

Research design

Barbell peak velocity during lifting was monitored in an observational research study. Specifically, a total of five devices recorded concentric peak velocity (see Figure 2): three LPTs (GYM, VIT, and GRB), one laser optic array, FLX, and one video-based mobile phone app, WLA, used in conjunction with an iPhone 13 Pro (Apple Inc., CA, USA) recording at 30 fps. These devices were chosen to reflect a range of popular devices currently used in the field of resistance training. Device use followed procedures recommended by each respective manufacturer. In previous VBT device validity studies, a three-dimensional motion capture system has been highly recommended as the gold standard criterion measure.^6,33 Thus, in this study, an OptiTrack s250e 12-camera motion capture system (Natural Point, Oregon, USA) was used as the criterion measure. This specific system is sufficiently reliable for biomechanical analyses compared to the Vicon M2 camera system⁴⁸ Furthermore, the accuracy of the OptiTrack system has been validated through comparisons with Vicon in previous research.⁴⁹ For WLA and the OptiTrack motion capture system, recordings were trimmed to only include the concentric, or pull phase, motion of the lift.

Subjects

The procedures and methods used in this study were approved by the University's Internal Review Board, IRB approval no. 20220760. All participants were informed of the study procedures, provided written consent prior to beginning the experiment, and given the option to withdraw at any time. Participants were recruited from local gyms in Broward, Hillsborough, and Miami-Dade counties through advertisement flyers posted in common areas on and surrounding the University's main campus, as well as Facebook, Instagram, and X (formerly Twitter) posts in local weightlifting groups. The total number of participants was 12 healthy competitive weightlifters (7 male and 5 female), aged between 21 and 50 years of age (29.4 ± 9.6 years). Mean participant height and weight were 1.7 ± 0.1 m and 77.3 ± 17.2 kg, respectively. Enrollment started on 17 November 2022 and ended on 15 April 2023. Inclusion criteria were no current or prior history of musculoskeletal injury or any medical conditions potentially affecting normal weightlifting kinematics; participation in at least two regional competitions; and a minimum weightlifting practice of three times per week. Participants had to be able to snatch and clean at least 80% and 120% of their body weight, respectively. Based on the inclusion criteria and in terms of weightlifting experience, participants were considered competitive or sub-elite athletes, which differ from those of elite lifters or resistance-trained athletes. Similarly to previous studies,⁵⁰ self-reported training 1RMs were attained within a time period of <3 months were considered (81.8 ± 22.5 kg snatch and 102.7 ± 25.9 kg clean).

Procedures

Motion capture system calibrations were performed prior to each data recording session and showed a negligible overall wand error and residual mean error of 0.35 mm and 0.70 mm, respectively. Calibration ratings in Optitrack (Poor, Fair, Good, Excellent, and Exceptional) are determined by the mean error. All 12 calibrations were done to an “excellent” or higher rating, as recommended by the manufacturer. A single 4 cm diameter spherical reflective marker was placed at the end of the barbell for displacement tracking, see Figure 3. Data were collected at 80 Hz. Each session was supervised by a USA Weightlifting Level 1 or higher certified coach to ensure proper lifting technique. After an initial warm-up, participants went through a sequence of single lifts with weight increasing from an unloaded barbell to 70% of their self-reported 1RM. Details on the warm-up routine and lift sequence are reported in Table 2. Typically, weightlifters train at 75% to 80% of 1RM or more.^7,51,52 However, it is recognized that the accuracy of the devices tested may vary with the magnitude of the velocity of the barbell. Accordingly, barbell loads ranging from 40% to 70% of 1RM were chosen to test both light and moderate loads, representing the speed-strength and the strength-speed phases, respectively.⁵³ This approach allows measuring fast and moderately fast barbell velocities, ensuring that the devices’ accuracy is assessed across a more comprehensive velocity spectrum. Participants performed both snatch and clean and were allowed to rest ad libitum between each repetition and set, as well as in between clean and snatch testing, progressing from 40% to 70% of 1RM at their own pace in terms of weight increment. The examination of the barbell velocity was only conducted for three repetitions of both the snatch and clean performed at 40% and 70% of 1RM. For each of the 12 total lifts measuring barbell velocity, all four devices were set up for recording (GYM, VIT, GRB, and FLX) along with the WLA app. The GYM, VIT, and GRB were connected to the barbell via their respective straps, and the FLX was capped onto the end of the barbell. The WLA app was set up through an iPhone 13 pro and followed the developer's recommendations: the camera was perpendicular to the floor via a tripod stand, videos were taken from the side angle (sagittal plane), and the tripod stand was 3 m away from the barbell. Barbell velocities were transmitted to their respective apps and documented for statistical analysis purposes. WLA data were collected at 30 fps.

Figure 3.

Experimental setup with single marker and velocity-based training (VBT) devices. A single spherical marker was attached to one end of the barbell with double-sided tape for motion tracking in addition to the other VBT devices. (A) and (B) Lateral view close-up. (C) and (D) Angled frontal view wide shot.

Table 2.

Experimental procedure.

	Exercise	Repetitions	Sets
Free warmup
Muscular activation warmup	Hip 90/90 rotation plus extension	×6 per side	1
	Thread the needle	×8 per side	1
	Elevated pigeon pose	60 s per side	1
	Cobra push-ups plus tall plank	×8 plus 30 s	2
	Prone PVC press	×10	2
	Unloaded barbell: pull, high pull, tall muscle snatch, press behind neck, overhead squat, press snatch, drop snatch	×5	2
Lifts	Snatch Clean	×2@barbell only 3 × 1@40%* ×2@50% ×2@60% 3 × 1@70%*	2

*Only the lifts at 40% and 70% of the self-reported one-repetition maximum were recorded.

Statistical analyses

In this research, all statistical analyses were conducted using Minitab (21.1.1, Minitab LLC, Pennsylvania, USA) and R (4.3.2, R Foundation for Statistical Computing, Vienna, Austria).⁵⁴ Significance was set at 95% (α = 0.05) for all tests. Data processing for the motion capture was conducted via Matlab R2023a (MathWorks Inc., Massachusetts, USA). Descriptive statistics were used to report peak velocity means, mean differences, and standard deviations. The mean difference represents the mean peak velocity difference between the gold standard (motion capture system) and the devices. Validity between the gold standard and the alternative methods (GYM, VIT, GRB, FLX, and WLA) was evaluated using least product regression.^55–60 Should the 95% confidence interval for the intercept exclude 0, it indicated the presence of fixed bias. Similarly, if the 95% confidence interval for the slope excluded 1, it indicated proportional bias. The existence of either fixed or proportional bias implied that the device was unsuitable for accurately predicting the motion capture peak velocity. Furthermore, a Bland–Altman analysis was employed to calculate the systematic bias,^61–64 represented by the mean difference for each device along with its 95% confidence interval limits. The 95% limits of agreement, defined as the systematic bias ± 1.96 times the standard deviation of the mean differences, were also determined. A systematic bias was deemed present if the range of the 95% confidence interval limits of the mean differences did not encompass the value 0. Reliability was determined from the magnitude of the standard error of measurement (SEM) and smallest detectable difference (SDD).^65,66

To determine the difference of each device in comparison to the gold standard motion capture system, we conducted an analysis of variance (ANOVA). To address the issue of missing values, which arose due to the devices not recording readings (106 missing values out of 720 records), a general linear model (GLM) was adopted for the ANOVA. Grubb's tests were performed to remove outliers within our dataset. The factors included in the ANOVA were the 1RM percentage (40% or 70%) and Device (GYM, VIT, GRB, FLX, and WLA). Additionally, the participant was included as a covariate in this analysis to control for potential confounding effects. Notably, all interaction effects were carefully examined in our analysis, and we utilized the backward elimination method to systematically remove non-significant factors and optimize the model while addressing errors as a response variable. To ensure that the assumptions of the ANOVA were met, the Levine test and Shapiro–Wilk test were employed to assess the homogeneity of variances (residuals) within each group and evaluate the normal distribution of residuals.

After the ANOVA, post hoc test, Tukey–Kramer test was conducted for pairwise comparisons to identify which groups were different. To quantify the corrected effect size of mean peak velocity differences across devices, we utilized a Hedges’ g test.^66,67 Finally, G*Power (version 3.1) was employed to calculate statistical power based on the sample size, considering the device and 1RM percentage as the main effects.

Results

The results of the least products regression analysis between the motion capture and the devices are presented in Table 3. The GYM was the only device that had no fixed or proportional bias. All other devices presented bias in at least one of the tested lift types and 1RM percentage. The Bland–Altman analysis did not show any significant systematic bias; confidence limits of the mean differences for all devices contained the value 0 (see Figure 4).

Figure 4.

Results of the Bland–Altman analysis. Bland–Altman plot with a mean difference between gold standard and devices. Limits of agreement were reported as systematic bias (mean difference) ± 1.96 × the standard deviation of the mean differences. (A) GymAware RS. (B) Vitruve. (C) GravityBox. (D) GymAware FLEX. (E) WL Analysis.

Table 3.

Results of validity least products regression analysis.

Devices	Lift	1RM%	Slope	95% confidence interval	Intercept	95% confidence interval
GYM	CL	40%	1.044	(0.910, 1.199)	−0.141	(−0.502, 0.171)
	CL	70%	1.058	(0.939, 1.193)	−0.189	(−0.461, 0.051)
	SN	40%	1.034	(0.870, 1.229)	−0.046	(−0.546, 0.372)
	SN	70%	0.945	(0.785, 1.136)	0.111	(−0.330, 0.481)
VIT	CL	40%	0.915	(0.845, 0.996) †	0.015	(−0.168, 0.174)
	CL	70%	1.166	(0.990, 1.378)	−0.498	(−0.919, −0.150) *
	SN	40%	0.814	(0.755, 0.877) †	0.307	(0.141, 0.464) *
	SN	70%	0.914	(0.811, 1.028)	0.039	(−0.223, 0.275)
GRB	CL	40%	1.090	(0.912, 1.307)	−0.297	(−0.765, 0.088)
	CL	70%	1.087	(0.827, 1.436)	−0.284	(−0.966, 0.222)
	SN	40%	1.492	(1.090, 2.131) †	−1.326	(−2.952, −0.304)*
	SN	70%	1.228	(0.961, 1.589)	−0.586	(−1.400, 0.015)
FLX	CL	40%	1.055	(0.833, 1.340)	−0.294	(−0.956, 0.222)
	CL	70%	1.856	(1.461, 2.446) †	−1.896	(−3.083, −1.100) *
	SN	40%	2.587	(1.713, 4.703) †	−4.066	(−9.419, −1.857) *
	SN	70%	1.312	(0.980, 1.800)	−0.831	(−1.953, −0.068) *
WLA	CL	40%	0.935	(0.807, 1.081)	0.007	(−0.327, 0.299)
	CL	70%	1.345	(1.139, 1.601) †	−0.898	(−1.406, −0.490) *
	SN	40%	1.015	(0.899, 1.147)	−0.110	(−0.457, 0.196)
	SN	70%	0.986	(0.784, 1.239)	−0.048	(−0.631, 0.417)

1RM%: one-repetition maximum percent.

Fixed bias is present (*) if the 95% confidence interval for the intercept does not include 0. Proportional bias is present if the 95% confidence interval for the slope does not include 1(†).

Mean peak velocity, mean difference, confidence intervals, SEM, SDD, and corrected effect size were determined for each lifting condition and are reported in Table 4. In Figure 5,⁶⁸ each device mean difference was represented as a percentage of the gold standard values for each lift condition. Notably, the smallest SEM, SDD, and Hedges’g were determined for the GYM during each lift type and 1RM combination. Post hoc power analysis indicated power was equal to or larger than 95% when inferring statistical differences in device-1RM percentage conditions across mean difference values.

Figure 5.

Mean difference as a percent of the gold standard for one-repetition maximum (1RM) percentage and device. Clean and snatch mean difference percentages across the five devices organized in terms of lift type and 1RM percentage.

Table 4.

Mean PVs and MD for 1RM% and device.

Devices	Lift	1RM%	PV ± SD (m/s)	MD ± SD (m/s)	95% CI	SEM (m/s)	SDD (m/s)	ES
GYM	CL	40%	2.3029 ± 0.3285	−0.0350 ± 0.1180	(−0.0783, 0.0083)	0.002	0.0055	0.079946
	CL	70%	1.9284 ± 0.1770	−0.0863 ± 0.0615	(−0.1089, −0.0638) ↓	0.002	0.0055	−0.32205
	SN	40%	2.5733 ± 0.2286	0.0444 ± 0.1158	(0.0011, 0.0877) ↑	0.003	0.0083	0.246539
	SN	70%	2.2368 ± 0.2756	−0.0737 ± 0.2124	(−0.1503, 0.0029)	0.001	0.0028	−0.32461
VIT	CL	40%	2.0974 ± 0.3161	−0.1730 ± 0.0775	(−0.1997, −0.1464) ↓	0.002	0.0055	−0.53826
	CL	70%	1.8123 ± 0.2056	−0.1769 ± 0.0732	(−0.2020, −0.1518) ↓	0.003	0.0083	−0.89362
	SN	40%	2.3697 ± 0.1758	−0.1602 ± 0.0734	(−0.1867, −0.1337) ↓	0.004	0.0111	−1.27423
	SN	70%	2.1444 ± 0.2177	−0.1587 ± 0.0742	(−0.1846, −0.1328) ↓	0.003	0.0083	−0.99272
GRB	CL	40%	2.0526 ± 0.2198	−0.1028 ± 0.0876	(−0.1374, −0.0681) ↓	0.003	0.0083	−0.77226
	CL	70%	1.8468 ± 0.1714	−0.1101 ± 0.0897	(−0.1499, −0.0703) ↓	0.004	0.0111	−0.78817
	SN	40%	2.4675 ± 0.2876	−0.0755 ± 0.1726	(−0.1484, −0.0026) ↓	0.022	0.0610	−0.31462
	SN	70%	2.1823 ± 0.1770	−0.0726 ± 0.0899	(−0.1089, −0.0363) ↓	0.003	0.0083	−0.91877
FLX	CL	40%	2.1766 ± 0.3462	−0.1631 ± 0.1910	(−0.2357, −0.0904) ↓	0.027	0.0748	−0.28646
	CL	70%	1.8572 ± 0.2654	−0.1536 ± 0.1632	(−0.2157, −0.0915) ↓	0.002	0.0055	−0.56449
	SN	40%	2.4577 ± 0.2459	−0.0840 ± 0.2008	(−0.1651, −0.0029) ↓	0.064	0.1774	−0.42237
	SN	70%	2.1136 ± 0.2300	−0.1109 ± 0.1942	(−0.1911, −0.0308) ↓	0.017	0.0471	−1.12584
WLA	CL	40%	2.1440 ± 0.3446	−0.1424 ± 0.1291	(−0.1906, −0.0942) ↓	0.006	0.0166	−0.3814
	CL	70%	1.7750 ± 0.2451	−0.2093 ± 0.1064	(−0.2477, −0.1710) ↓	0.004	0.0111	−0.97456
	SN	40%	2.4627 ± 0.2048	−0.0699 ± 0.1332	(−0.1171, −0.0227) ↓	0.004	0.0111	−0.47145
	SN	70%	2.2247 ± 0.2414	−0.0799 ± 0.1309	(−0.1271, −0.0327) ↓	0.007	0.0194	−0.43867

PV: peak velocity; SD: standard deviation; MD: mean difference; 1RM%: one-repetition maximum percentage; CI: confidence interval; SEM: standard error of measurement; SDD: smallest detectable difference; ES: effect size; CL: clean; SN: snatch.

In the 95% CI column, (↓) signified underestimation and (↑) signified overestimation of the device compared to the motion capture system, while a lack of symbol signified no bias. ES per Hedge's g test was also reported, where 0.2 represented a small effect size, 0.5 was a medium effect size, and 0.8 was a large effect size.

Based on the GLM analysis performed, the interaction between the 1RM percentage and the VBT device had a significant effect for the clean (p = 0.035) but not for the snatch (p = 0.231). For both lifts, the errors associated with the devices investigated were statistically different (p < 0.001). Mean marginal values and significant groupings based on Tukey–Kramer pairwise comparison are reported in Table 5. For the clean, the devices performed as follows, ordered by decreasing error: WLA (Δ = −0.178 m/s), VIT (Δ = −0.175 m/s), FLX (Δ = −0.146 m/s), GRB (Δ = −0.112 m/s), and GYM (Δ = −0.060 m/s). For the snatch, the devices performed as follows, ordered by decreasing error: VIT (Δ = −0.171 m/s), FLX (Δ = −0.121 m/s), WLA (Δ = −0.075 m/s), GRB (Δ = −0.060 m/s), and GYM (Δ = 0.012 m/s).

Table 5.

Tukey mean differences from motion capture peak velocity for 1RM% and device.

Lift type	Variable		Tukey mean difference (Δ m/s)	Tukey groupings	ANOVA p-value
Clean	1RM %	40%	−0.120	A	0.035
	1RM %	70%	−0.148	B	0.035
	Device	GYM	−0.060	A	<0.001
		GRB	−0.112	A, B
		FLX	−0.146	B, C
		VIT	−0.175	C
		WLA	−0.178	C
Snatch	1RM%	40%	−0.075	A	0.231
	1RM%	70%	−0.091	A	0.231
	Device	GYM	0.012	A	<0.001
		GRB	−0.060	B
		WLA	−0.075	B
		FLX	−0.121	B, C
		VIT	−0.171	C

1RM%: one-repetition maximum percentage; ANOVA: analysis of variance.

Negative values indicated underestimated velocity and a positive value reflected overestimated velocity. The marginal means for 1RM% and device were grouped using a Tukey–Kramer pairwise comparison.

Discussion

To validate the accuracy of five devices that measure barbell velocity (GYM, VIT, GRB, FLX, and WLA), the present study compared the peak velocity during the concentric phase of the snatch and clean to those determined using a gold standard motion capture system. To the authors’ best knowledge, this is the first singular study simultaneously validating and comparing these five specific devices currently trending in the market, with respect to both clean and snatch.

In terms of validity and reliability, the only measurements that showed no fixed and proportional bias were for the GYM. All other devices had two or more combinations showing bias. In addition, the smallest SEM and SDD were found in GYM for each type of lift and 1RM combination (see Table 4), suggesting more consistent and precise measurements compared to other devices when assessing peak velocity in clean and snatch. In addition, while the Bland–Altman plots did not show any significant systematic bias, GYM had the smallest mean difference (0.023 m/s) and limits of agreement (0.0226 and −0.181 m/s) with respect to the gold standard. Notably, all devices had a small SEM and SDD compared to the mean difference, exhibiting favorable characteristics regarding measurement precision.

When comparing the mean differences of each device to the gold standard, four out of the five devices significantly differ in assessing peak velocity during weightlifting (see Table 4). VIT, FLX, and GRB all underestimated the velocity values for clean at 40% and 70%: mean differences were 7.62% and 8.89% (0.17 and 0.18 m/s), 6.97% and 7.64% (0.16 and 0.15 m/s), and 4.77% and 5.63% (0.10 and 0.11 m/s), respectively (see Figure 5). This significant underestimation pattern continued for the snatch at 40% and 70% with errors of 6.33% and 6.89% (0.16 m/s for both) for VIT, 3.30% and 4.99% (0.08 and 0.11 m/s) for FLX, and 2.97% and 3.22% (0.08 and 0.07 m/s) for GRB. The bias in these LPT devices may suggest an inherent error in their encoders (i.e. not accounting for horizontal displacement and/or low sampling rate), particularly when put to the test with the high-velocity ballistic movements of the barbell performed in the clean and snatch.⁶

The app WLA significantly underestimated peak velocity with a mean difference of 6.23% and 10.55% (0.14 and 0.21 m/s) for the 40% and 70% clean trials and 2.76% and 3.47% (0.07 and 0.08 m/s) during the snatch, respectively. The discrepancy in each of these exercises can be explained by the app's limited sampling rate compared to the motion capture system. Specifically, the WLA app sampled at only 30 fps at the time of data collection, while the motion capture was set at 80 fps. Furthermore, some mobile devices can record at a higher sampling rate. For example, the iPhone 14 can record up to 60 fps at normal speed, and 240 fps in ultra-slow motion. As a result of this difference, the motion capture system collects more coordinate points across the lift, resulting in a larger sample size, and therefore it is more likely to capture higher velocity points. Only the GYM was not significantly different from the criterion measure in two of the four testing conditions: 40% 1RM clean and 70% 1RM snatch. For these conditions of 40 and 70%, the mean difference was only 1.50% (0.035 m/s) and 3.19% (0.074 m/s), respectively. Comparatively, the other two GYM conditions (70% 1RM clean and 40% 1RM snatch) had a mean difference of <0.090 m/s, significantly differing, but having a lower variation to the other devices. Previous studies have shown GYM, the most expensive LPT on the market, to be valid in its peak velocity values, even for ballistic movements.^33–36 Notably, the smallest Hedges’ g test results were found in GYM (Table 4), supporting the Tukey–Kramer pairwise comparison results and suggesting that there is not a meaningful difference between GYM when compared to the motion capture system.

For both the clean and the snatch, there were significant differences when values from the devices were compared with each other. As seen in Table 5, for the clean, the WLA, VIT, and FLX performed equally poorly with a mean difference of approximately 0.17 m/s. These devices are relatively inexpensive compared to other market options, and previous research shows them to be unvalidated with lifts contested in the sport of weightlifting.^33,39,42 Interestingly, the second-best device for monitoring velocity during the clean was GRB with an error of 0.112 m/s, despite it being the least expensive LPT tested. As expected, the GYM was the best device for monitoring velocity during the clean with a mean difference of only 0.060 m/s. For the snatch, the VIT had the highest mean difference of 0.171 m/s. Next, FLX was the third-best snatch device with a mean difference of 0.121 m/s. WLA and GRB were tied as the second-best option with a mean difference of 0.075 and 0.060 m/s, respectively. While GRB ranked in the same position as the clean, WLA rose in its performance with the snatch as compared to the clean. Again, GYM was the best-performing snatch device with an error of 0.012 m/s. We expected that LPTs (GYM, VIT, and GRB) would show higher validity compared to the laser optic array (FLX) due to their higher sampling rate. WLA appears to exhibit a level of validity comparable to most of the devices, except for GYM. Consequently, the validation results for GYM are consistent with prior research^33–36 and our hypothesis, considering its higher retail price compared to the other devices examined in this study. In addition, the GLM showed that the errors associated with the 1RM percentages of 40 and 70 were significantly different for the clean, but not for the snatch. While the snatch had an overall smaller error compared to the clean, the difference between 40% and 70% RM percentage error was larger in the clean than the snatch, suggesting that these devices cannot capture high peak velocities precisely. Overall, both 1RM and lift type differences suggest that these devices tend to be biased when recording high-velocity ballistic movements seen in the sport of weightlifting due to limited sample size when recording barbell velocity.

The performance of the barbell velocity devices in the sport of weightlifting strongly affects the accuracy of the 1RM prediction. This 1RM prediction is calculated through the measurement of displacement, and that measure is differentiated to estimate peak concentric velocity at submaximal percentages of 1RM. If the peak velocity is consistently underestimated (as happens with all devices except the GYM based on our results), then the 1RM calculation will also be underestimated. This means that the potential training programs would be based on a lower prediction of the true 1RM, which would not enhance the performance of the lifter to its maximum potential. In addition, based on the bias analysis, if the 40% peak velocity is underestimated, but the 70% is not, then the slope of the line would result in an overestimation of the 1RM. Therefore, practitioners must consider the errors and biases of each device before implementing VBT. As shown, there are multiple device options on the market that provide different combinations of price and accuracy.

A limitation of the study is that only 40% and 70% of 1RM were tested, and the GLM showed that 1RM percentage was not a significant variable impacting velocity accuracy for the snatch. To enhance validity, a broader range of subpercentages of 1RM would have been beneficial. However, while participants executed lifts at 50% and 60% of 1RM, we opted not to capture velocity measurements at these points, as they were deemed to provide negligible additional value to the assessment. Another limitation is related to the level of proficiency in the clean and snatch techniques, which in this study is representative of a national and regional competitive athletes’ population. This level of proficiency may differ from those of elite lifters or resistance-trained athletes. Future studies should be conducted to examine populations of athletes with different levels of proficiency in clean and snatch techniques. Furthermore, Rossi et al.⁶⁹ reported a possible 5 cm difference between the right and left side of the barbell during the snatch. To account for such a difference, Lake et al.⁵⁰ suggest the use of markers on both ends of the barbell. This approach would certainly be the best practice, especially when measuring average barbell velocities. In our study, we placed a single marker on one end of the barbell. However, we do not believe this approach may have led to measurable errors, given the specific measurement we focused on. In fact, the present study evaluated the instantaneous peak velocity of the barbell. This quantity is calculated by determining the difference in the barbell position between two consecutive sampling points and dividing this value by the time interval between those two consecutive sampling points. Given the short interval of time between two consecutive sampling points (∼0.01 s), it would be reasonable to assume that the barbell moves in the space as a rigid body (i.e. negligible deformations of the barbell). Consequently, any difference in the displacements of the two ends of the barbell would be negligible. Therefore, we believe that the omission of the reflective marker on the side of the FLX and GRB for the specific velocity measurement conducted in the present study does not introduce a significant source of systematic experimental error. Finally, only the clean and the snatch were analyzed in the present study. An assessment of these five barbell velocity devices' precision could also be expanded to the field of powerlifting, which traditionally consists of the deadlift, back squat, and bench press, lift types typically slower in terms of peak barbell velocity compared to weightlifting. This would warrant a separate study.

In summary, this study examined the accuracy of peak velocities recorded by three LPTs, one laser optic array, and one video-based app in the sport of weightlifting. Our findings indicated that GYM is the superior option among barbell velocity devices in the sport of weightlifting for competitive athletes. If a lifter is looking for a more economical alternative to GYM that still has relatively high accuracy, the researchers would recommend GRB. Moreover, a lifter's ultimate choice of device would depend on the level of velocity error that is acceptable to them in their personal implementation of VBT.

Practical applications

Practitioners deciding to utilize VBT must consider both accuracy and cost when determining a device to purchase for use in their program. While some units were significantly cheaper, roughly one-eighth the cost of GYM, the accuracy and reliability of the device suffered. As every situation is different, there is no single best device when both price and accuracy are concerned. A more expensive unit is generally more accurate, and a cheaper unit is generally less accurate. The tradeoff between cost and accuracy must be decided by the practitioner. Based on the results of this study, the most accurate device was GYM. Therefore, practitioners should choose this as their first option. However, factoring in cost, if they decide to use other devices, they can use the results of the current study to compare the tradeoff of error percentage to 1RM prediction.

Footnotes

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Cheng-Bang Chen

Francesco Travascio

References

Newton

. Weightlifting? Weight lifting? Olympic lifting? Olympic weightlifting? Strength Cond J 1999; 21: 15.

Suchomel

Nimphius

Bellon

, et al. The importance of muscular strength: Training considerations. Sports Med 2018; 48: 765–785.

Suchomel

Nimphius

Stone

. The importance of muscular strength in athletic performance. Sports Med 2016; 46: 1419–1449.

Weakley

Mann

Banyard

, et al. Velocity-based training: from theory to application. Strength Cond J 2021; 43: 31–49.

Haff

Triplett

. Essentials of strength training and conditioning. 4th ed. Colorado Springs, CO: Human Kinetics, 2015.

Weakley

Morrison

García-Ramos

, et al. The validity and reliability of commercially available resistance training monitoring devices: A systematic review. Sports Med 2021; 51: 443–502.

Garhammer

Takano

. Training for weightlifting. In: Komi P (ed) Strength and power in sport. 2nd ed. Hoboken, NJ, 2003, pp.502–515.

Berton

Soriano

da Silva

, et al. Concurrent validity and reliability of the load–velocity relationship to predict the one-repetition maximum during three weightlifting derivatives. Kinesiology 2021; 53: 215–225.

Picerno

Iannetta

Comotto

, et al. 1RM Prediction: A novel methodology based on the force-velocity and load-velocity relationships. Eur J Appl Physiol 2016; 116: 2035–2043.

10.

Garcia-Ramos

Barboza-Gonzalez

Ulloa-Diaz

, et al. Reliability and validity of different methods of estimating the one-repetition maximum during the free-weight prone bench pull exercise. J Sports Sci 2019; 37: 2205–2212.

11.

Pérez-Castilla

Jerez-Mayorga

Martínez-García

, et al. Comparison of the bench press one-repetition maximum obtained by different procedures: Direct assessment vs. lifts-to-failure equations vs. two-point method. Int J Sports Sci Coach 2020; 15: 337–346.

12.

Abadjiev

. Preparation of the Bulgarian weightlifters for the Olympic Games 1984. Aust Weightlifter 1981: 25–29.

13.

Abadjiev

. Basic training principles for Bulgarian elite. Inter Olympic Lifter 1976; 3: 12–13.

14.

Harris

Cronin

Taylor

K-L

, et al. Understanding position transducer technology for strength and conditioning practitioners. Strength Cond J 2010; 32: 66–79.

15.

Jovanović

Flanagan

. Researched applications of velocity based strength training. J Aust Strength Cond 2014; 22: 58–69.

16.

Thompson

Olusoga

Rogerson

, et al. Is it a slow day or a go day?”: The perceptions and applications of velocity-based training within elite strength and conditioning. Int J Sports Sci Coach 2023; 18: 1217–1228.

17.

Sánchez-Medina

González-Badillo

. Velocity loss as an indicator of neuromuscular fatigue during resistance training. Med Sci Sports Exerc 2011; 43: 1725–1734.

18.

Harbili

Alptekin

. Comparative kinematic analysis of the snatch lifts in elite male adolescent weightlifters. J Sports Sci Med 2014; 13: 417–422.

19.

Weakley

Cowley

Schoenfeld

, et al. The effect of feedback on resistance training performance and adaptations: A systematic review and meta-analysis. Sports Med 2023; 53: 1789–1803.

20.

Randell

Cronin

Keogh

, et al. Effect of instantaneous performance feedback during 6 weeks of velocity-based resistance training on sport-specific performance tests. J Strength Cond Res 2011; 25: 87–93.

21.

Thompson

Rogerson

Ruddock

, et al. Pooled versus individualized load-velocity profiling in the free-weight back squat and power clean. Int J Sports Physiol Perform 2021; 16: 825–833.

22.

Haff

Garcia-Ramos

James

. Using velocity to predict the maximum dynamic strength in the power clean. Sports (Basel) 2020; 8: 1–12.

23.

Hughes

Banyard

Dempsey

, et al. Using a load-velocity relationship to predict one repetition maximum in free-weight exercise: A comparison of the different methods. J Strength Cond Res 2019; 33: 2409–2419.

24.

Banyard

Nosaka

Haff

. Reliability and validity of the load–velocity relationship to predict the 1RM back squat. J Strength Cond Res 2017; 31: 1897–1904.

25.

Fry

Kraemer

Lynch

, et al.

Does short-term near-maximal intensity machine resistance training induce overtraining?

J Strength Cond Res 1994; 8: 188–191.

26.

Blumert

Crum

Ernsting

, et al. The acute effects of twenty-four hours of sleep loss on the performance of national-caliber male collegiate weightlifters. J Strength Cond Res 2007; 21: 1146–1154.

27.

Zourdos

Dolan

Quiles

, et al. Efficacy of daily one-repetition maximum training in well-trained powerlifters and weightlifters: a case series. Nutr Hosp 2016; 33: 437–443.

28.

Guppy

Nagatani

Poon

WCK

, et al. The stability of the deadlift three repetition maximum. Int J Sports Sci Coach 2024; 19: 812–821.

29.

Guppy

Nagatani

Poon

WCK

, et al. Changes in deadlift six repetition maximum, countermovement jump performance, barbell velocity, and perceived exertion over the duration of a microcycle. Int J Sports Sci Coach 2024; 19: 822–831.

30.

Janicijevic

García-Ramos

Lamas-Cepero

, et al. Comparison of the two most commonly used gold-standard velocity monitoring devices (GymAware and T-force) to assess lifting velocity during the free-weight barbell back squat exercise. Proc IMechE Part P: J Sports and Engineering Technology 2021: 237.

31.

Weakley

Munteanu

Cowley

, et al. The criterion validity and between-day reliability of the perch for measuring barbell velocity during commonly used resistance training exercises. J Strength Cond Res 2023; 37: 787–792.

32.

GymAware. https://shop.gymaware.com/?target=international (2021, accessed 7 November 2023).

33.

Fritschi

Seiler

Gross

. Validity and effects of placement of velocity-based training devices. Sports (Basel) 2021; 9: 123.

34.

Banyard

Nosaka

Sato

, et al. Validity of Various methods for determining velocity, force, and power in the back squat. Int J Sports Physiol Perform 2017; 12: 1170–1176.

35.

Dorrell

Moore

Smith

, et al. Validity and reliability of a linear positional transducer across commonly practised resistance training exercises. J Sports Sci 2019; 37: 67–73.

36.

Mitter

Hölbling

Bauer

, et al. Concurrent validity of field-based diagnostic technology monitoring movement velocity in powerlifting exercises. J Strength Cond Res 2021; 35: 2170–2178.

37.

Thompson

Rogerson

Dorrell

, et al. The reliability and validity of current technologies for measuring barbell velocity in the free-weight back squat and power clean. Sports (Basel) 2020; 8: 94.

38.

Vitruve. https://shop.vitruve.fit/ (2020, accessed 7 November 2023).

39.

Kilgallon

Cushion

Joffe

, et al. Reliability and validity of velocity measures and regression methods to predict maximal strength ability in the back-squat using a novel linear position transducer. Proc IMechE Part P: J Sports Engineering and Technology 2022; 0.

40.

Gravity Box. https://www.gravitybox.ca/product-page (2022, accessed 7 November 2023).

41.

GymAware. https://www.flexstronger.com/product/flex/ (2019, accessed 7 November 2023).

42.

Weakley

Chalkley

Johnston

, et al. Criterion validity, and interunit and between-day reliability of the FLEX for measuring barbell velocity during commonly used resistance training exercises. J Strength Cond Res 2020; 34: 1519–1524.

43.

Sánchez-Pay

Courel-Ibáñez

Martínez-Cava

, et al.

Is the high-speed camera-based method a plausible option for bar velocity assessment during resistance training?

Measurement (Lond) 2019; 137: 355–361.

44.

WL Analysis. https://wlanalysis.com/#page-top (2021, accessed 7 November 2023).

45.

Courel-Ibáñez

Martínez-Cava

Morán-Navarro

, et al. Reproducibility and repeatability of five different technologies for bar velocity measurement in resistance training. Ann Biomed Eng 2019; 47: 1523–1538.

46.

Balsalobre-Fernández

Marchante

Muñoz-López

, et al. Validity and reliability of a novel iPhone app for the measurement of barbell velocity and 1RM on the bench-press exercise. J Sports Sci 2018; 36: 64–70.

47.

Martínez-Cava

Hernández-Belmonte

Courel-Ibáñez

, et al. Reliability of technologies to measure the barbell velocity: implications for monitoring resistance training. PloS one 2020; 15: e0232465.

48.

Hansen

Gibas

Honeine

J-L

, et al. An inexpensive solution for motion analysis. Proc IMechE Part P: J Sports Engineering and Technology 2014; 228: 165–170.

49.

Nagymate

Kiss

. Application of OptiTrack motion capture systems in human movement analysis A systematic literature review. Recent Innov Mechatron 2018: 5. DOI: https://doi.org/10.17667/riim.2018.1/13

50.

Lake

Augustus

Austin

, et al. The reliability and validity of the bar-mounted PUSH BandTM 2.0 during bench press with moderate and heavy loads. J Sports Sci 2019; 37: 2685–2690.

51.

Ajan

Baroga

. Weightlifting: Fitness for all sports. Budapest: International Weightlifting Federation, 1988.

52.

Takano

. Weightlifting programming: A winning coach's guide. Sunnyvale, CA, USA: Catalyst Athletics, Inc., 2012, 246 p.

53.

Comfort

Haff

Suchomel

, et al. National Strength and Conditioning Association position statement on weightlifting for sports performance. J Strength Cond Res 2023; 37: 1163–1190.

54.

R-CoreTeam. A Language and Environment for Statistical Computing. Vienna, Austria: R Foundation for Statistical Computing, 2023, https://www.R-project.org/ .

55.

Mullineaux

Barnes

Batterham

. Assessment of bias in comparing measurements: a reliability example. Meas Phys Educ Exerc Sci 1999; 3: 195–205.

56.

Ludbrook

. Comparing methods of measurements. Clin Exp Pharmacol Physiol 1997; 24: 193–203.

57.

Ludbrook

. Statistical techniques for comparing measurers and methods of measurement: a critical review. Clin Exp Pharmacol Physiol 2002; 29: 527–536.

58.

Ludbrook

. Linear regression analysis for comparing two measurers or methods of measurement: but which regression? Clin Exp Pharmacol Physiol 2010; 37: 692–699.

59.

Bland

Altman

. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet 1986; 327: 307–310.

60.

Bland

Altman

. Agreement between methods of measurement with multiple observations per individual. J Biopharm Stat 2007; 17: 571–582.

61.

Giavarina

. Understanding Bland Altman analysis. Biochem Med (Zagreb) 2015; 25: 141–151.

62.

Goldsmith

Trepeck

Halle

, et al. Validity of the open barbell and Tendo weightlifting analyzer systems versus the Optotrak Certus 3D motion-capture system for barbell velocity. Int J Sports Physiol Perform 2019; 14: 540–543.

63.

Nagao

Yamashita

. Validation of video analysis of marker-less barbell auto-tracking in weightlifting. PLOS ONE 2022; 17: e0263224.

64.

Sandau

Chaabene

Granacher

. Concurrent validity of barbell force measured from video-based barbell kinematics during the snatch in male elite weightlifters. PLOS ONE 2021; 16: e0254705.

65.

Beckerman

Roebroeck

Lankhorst

, et al. Smallest real difference, a link between reproducibility and responsiveness. Qual Life Res 2001; 10: 571–578.

66.

Bernards

Sato

Haff

, et al. Current research and statistical practices in sport science and a need for change. Sports (Basel) 2017; 5: 87.

67.

Lakens

. Calculating and reporting effect sizes to facilitate cumulative science: A practical primer for t-tests and ANOVAs. Front Psychol 2013; 4: 863.

68.

Weissgerber

Milic

Winham

, et al. Beyond bar and line graphs: time for a new data presentation paradigm. PLoS Biol 2015; 13: e1002128.

69.

Rossi

Buford

Smith

, et al. Bilateral comparison of barbell kinetics and kinematics during a weightlifting competition. Int J Sports Physiol Perform 2007; 2: 150–158.