Extending the curve: A Closer Look at High-Velocity Measures in the Power Clean.

Wei Qian Lim1, MS, David Smith2, Eric D. Magrum2, PhD,
The George Washington University1, Washington, DC
James Madison University2, Harrisonburg, Virginia

Editor’s Note: Table 1 was incorrectly published. This has been corrected. Tables 2 and 3 were reformatted during the revision process.

Abstract
Purpose: This study examines the validity and reliability of a commercially available velocity-based training device, GymAware, when measuring barbell velocity during submaximal power cleans. While GymAware has been validated for slower movements, limited research has assessed its accuracy at higher velocities, particularly in Olympic weightlifting derivatives.
Methods: Ten resistance-trained participants completed two sets of five repetitions at 40%, 50%, and 60% of their perceived one-repetition maximum in the power clean. Mean and peak barbell velocity were recorded using GymAware and compared to a motion capture system as the criterion measure. Data were analyzed for reliability using intraclass correlation coefficients and validity through correlation and regression analysis.

Results: Mean velocity measurements from GymAware demonstrated strong agreement with motion capture across all loads, with correlations exceeding 0.85 and an intraclass correlation coefficient of 0.85, indicating good reliability. However, peak velocity measurements exhibited greater variability, with a systematic overestimation of 0.37 m/s and a lower reliability coefficient (0.31). Linear regression models confirmed that GymAware accounted for 88% of the variance in mean velocity but only 44% in peak velocity, suggesting less precision in high-velocity movements.

Conclusion: GymAware provides reliable and valid measurements of mean barbell velocity but has limitations in accurately assessing peak velocity during rapid weightlifting movements. Coaches and practitioners should prioritize mean velocity when utilizing velocity-based training for performance monitoring.

Application in Sports: Velocity-based training offers an efficient method for tracking performance and adjusting training loads. GymAware’s ability to measure mean velocity reliably makes it a useful tool for monitoring training adaptations and providing immediate feedback to athletes. However, practitioners should be cautious when interpreting peak velocity data, particularly in high-velocity Olympic weightlifting derivatives, and consider alternative methods for precise assessment.

Introduction
Resistance training is a well-documented modality for improving force production, power, lean body mass, and overall athletic performance (10-11,13,20,27). For these reasons resistance training has become synonymous with athlete preparation. Before the technological renaissance, tracking athletes’ progress and assessing program effectiveness was almost entirely comprised of assessing progressive overload via number of repetitions completed or through the manipulation of external load lifted (15,19,22). However, these more traditional methods come with several challenges, making it difficult to assess program effectiveness. Specifically, athlete’s perceived exertion, range of motion, and different pacing strategies can confound practitioners’ ability to assess meaningful changes as it relates to physiological adaptations resultant resulting from training (12,18,19,22). Because of this, numerous efforts have been made to leverage technological tools to enhance the assessment of training efficacy.

Recent technological advancements have popularized the tracking of barbell velocity, termed velocity-based training (VBT), and highlighted its usefulness in gauging training efficacy. VBT is utilized for a multitude of reasons, including but not limited to predicting 1 repetition maximum (1RM) without the accumulation of excessive fatigue and increased risk of injury, monitoring training performance and neuromuscular fatigue, and providing immediate kinematic feedback potentially leading to enhanced training outcomes (1,3,-4,7,-8,10-11,18,23,24,26,28). As with any technological tool, measures of validity and reliability are paramount to assess the meaningfulness of the data provided. Providing reliable data is important for coaches and athletes alike, to accurately assess the physiological changes associated with training programs, as well as make appropriate alterations when needed.

For over 20 years, GymAware (GYM) has been considered the gold standard of linear positional transducers (LPT). LPT’s function by measuring displacement of a barbell as well as the time taken to complete said displacement. By using this data, the LPT computes several variations of barbell velocity and power (average, peak, etc.) (17). Previous research suggests that the GYM is both highly valid and reliable at slow velocities (0.3-0.7 m/s) . (3-5,7-9,14,15,21). However, few studies have examined the reliability and validity of the GYM during low load, high velocity weightlifting or plyometric movements (0.7+ m/s). Studies that have investigated GYM at these velocities report that the GYM system typically underreports peak velocity and power outputs at lower loads and higher velocity (2,6,14).

Askow et al. (2) examined the reliability and validity of GYM software at both 60 and 80% of 1RM back squats. They found that GYM tends to underestimate peak velocity by 11.6% and software is not the most accurate measure of barbell velocity during high velocity movements. Despite this, Askow and his team of researchers still reported high levels of reliability at high velocities (2). Orange et al. (17) reported excellent reliability for both peak and mean velocity measurements at a range of different percentages of 1RM in the back squat and bench press with interclass correlations (ICCs) ranging from 0.96 to 0.99. Lorenzetti et al. (14) found that GYM was both reliable and valid at tracking bar velocity at 70% of 1RM and during a ballistic jump squats; however, they found much higher reliability and validity at lower velocities when compared to the high velocity jump squat plyometrics. A systematic review of LPTs and linear velocity transducers (LVT) corroborated these findings and reported that LPTs, including the GYM, were valid and reliable in measuring velocity during powerlifting and weightlifting movements . (25).
Another review on the subject highlights the need for independent investigations of velocity-based sensors to examine higher velocity lifts such as Olympic weightlifting derivatives (1.2-1.6 m/s) (16). Due to their unique utility and force-velocity characteristics, weightlifting movements , such as the snatch, clean and jerk, are routinely utilized in sport performance settings around the globe. An essential element of these lifts is how fast the weight moves. Few studies have compared such devices to a criterion measure, namely motion capture (25). However, existing research on devices like the GYM Power Tool suggests high validity and reliability when measuring velocity during high-velocity barbell movements. Orange et al. (17) reported excellent reliability of GYM for back squats and bench presses, with ICCs ranging from 0.96 to 0.99 for velocity, suggesting that it could similarly perform well in more dynamic lifts. There is limited research on the reliability and validity of LPDT when measuring velocity during Olympic lift derivatives. Thus, the current study will address the gap in the literature and extend our understanding of the validity and reliability of VBT devices at higher velocities. Specifically, the purpose of this study is to examine the reliability and validity of GYM compared to Qualisys Motion Capture during the power clean.

Methods
The study was carried out with 10 participants (Table 1). Participants had at least one year of prior experience strength training, defined as an average of two training sessions per week. Subjects were between the ages of 18-40, technically proficient in the clean, not pregnant, free of known cardiovascular, metabolic, or renal disease, and free of injuries. After giving written consent, technical proficiency in the clean was determined during a familiarization session prior to data collection.

Table 1. Participant Characteristics 

Sex	Age (years) (mean ± SD)	Height (m) (mean ± SD)	Weight (kg) (mean ± SD)	Predicted 1RM (kg) (mean ± SD)
Male (n=5)	23.4 ± 4.4	1.74 ± 0.06	83.3 ± 9.8	106.6 ± 24.5
Female (n=5)	22.0 ± 0.7	1.62 ± 0.06	72.6 ± 22.6	61.2 ± 17.0
Total (n=10)	22.7 ± 3.1	1.68 ± 0.09	78.0 ± 17.4	83.9 ± 31.1

For a clean, participants had to lift the barbell in one smooth move from the floor, catching the barbell in a front rack position. Feet were to be shoulder width apart or just outside shoulder width at the catch. The participants were cued to move the weight as quickly as possible while staying under control. Participants with working weights lighter than what could be provided with bumper plates, the lift began from a hang at mid-shin height.

During the familiarization session participants were asked to complete a health history questionnaire before height and weight were taken. After a general warm up that consisted of 50 jumping jacks, 10 bodyweight squats, 5 jump squats and 5 cleans with the empty barbell, the participants provided a perceived 1RM (ex. 200 lbs.). 50% of the participants’ perceived 1RM was loaded onto the barbell (ex. 50% of 200 lbs. = 100 lbs.). The participant was then asked to perform 1 set of 5 repetitions, at which point the research team determined if technical proficiency was sufficient (binary yes or no).

Participants who met the inclusion criteria and demonstrated proficiency in the clean were invited back for a lifting session. The session began with the same general warm-up detailed above. Participants whose schedules permitted both sessions to be completed consecutively (familiarization + lifting) were not asked to perform the warmup prior to the lifting session. In total, participants completed six sets: two sets of five repetitions at 40%, 50%, and 60% of perceived 1RM (ex. 200lbs 1RM: 40% = 80lbs, 50% = 100 lbs., and 60% = 120lbs). Each set began with the signal “You may begin your lift.” Participants were instructed to fully stop and/or set down the bar at the end of each repetition for at least a one count to prevent the use of momentum and allow for a distinct ending to each repetition. This was reinforced with a count of “one” between each repetition. Participants were given three minutes to rest between each set.
Qualisys motion capture system was used as a gold standard/criterion reference. The motion capture set-up consisted of six cameras: three from the Miqus M3 series and three from the Oqus series. Six reflective markers were attached to the barbell. Two markers were attached to either end of the bar, while four markers were attached in square configuration on the collar of the barbell (Figures 1 and 2). The data were recorded with the software QTM 2020-2 Build 5710, with a frequency of 100 Hz. The limits for standard deviation for wand length calibration were 0.3 and 0.5 mm.

The GYM RS, placed on the ground between the pad and platform, was tethered to the shaft of the barbell close to the four reflective markers (see Figures 1 and 3). The GYM RS device was connected via Bluetooth to the free version of the GYM iOS application (Version 4.0.1). GYM RS records at 50 Hz. Peak and mean velocity (m/s) for each repetition were hand recorded from the application into a Microsoft Excel spreadsheet.

Velocity data were exported from the Qualisys Track Manager (QTM) software to Microsoft Excel. The beginning of the lift was determined by the inflection of barbell velocity denoted by an increase of 0.01 m/s for three consecutive frames. The end of the concentric portion of the lift was determined by the first maximum velocity value or crest of velocity curve. Corresponding with GYM, mean concentric velocity (m/s) was determined by averaging marker velocities over the entire concentric portion of the lift. Peak concentric velocity (m/s) was calculated by averaging the individual velocities of each marker over a sample period of 20 milliseconds immediately preceding peak velocity.

Participants stood on a wooden platform with the barbell resting on black foam pads on either side of the platform. Unless the participant’s working weight utilized change plates or the empty bar, the clean started from the black foam pads. If not, the clean started from a hang at mid-shin height. The materials were a 20 kg bar, Rouge change plates between 0.5 and 5 kg, 2.5 and 5 lb. plates, as well as 25 and 45 lb. bumper plates. Working weights for each participant were calculated to get as close to 40%, 50%, and 60% of perceived 1RM.

Results
Data was collected for 10 participants during a single data collection session. Subjects completed six sets: two sets of five repetitions at 40%, 50%, and 60% of perceived 1RM. Mean and peak velocity was recorded using GYM and Qualisys motion capture software for each repetition. There was a total of 60 data points per participant, resulting in 600 total data points.
3.1 Validity

Figure 4. Scatter plots expressing the peak and mean bar velocities at 40, 50, and 60% of one repetition maximum as measured by GYM and Qualisys motion capture systems. Error is defined as the difference between the GYM measurements and Qualisys measurements, with cooler colors representing less error and hotter colors representing more error. Dashed line represents a perfect linear fit that assumes no variance between the two devices. All correlations were statistically significant with a p<0.05

Scatter plots for peak velocity at each percentage of 1RM showed varied levels of correlation between GYM and Qualisys. At 40% of 1RM r=0.706, at 50% r=0.512, and at 60% r=0.703. Each of the aforementioned correlations reached statistical significance at the 0.05 level and indicate a moderate correlation between the GYM and Qualisys measurements of bar velocity. 50% of 1RM demonstrated the highest variability (Figure 4).

The mean velocity measurements between the two systems demonstrated stronger correlations across all load percentages. At 40% r=0.958, at 50% r=0.938, and at 60% r=0.871. All correlations were statistically significant (p<0.05) and indicate a consistent, strong relationship between GYM and Qualisys when assessing mean bar velocity (Figure 4).
GYM software tended to overpredict peak barbell velocities at all intensities by 0.37 m/s on average, while only over predicting mean barbell velocity by 0.09 m/s (Figure 5).

Table 2. Comparison of Linear Regression Model Results for GYM and Qualisys Motion Capture System at Different Percentages of Perceived One Repetition Max

Load (%1RM)	R2		F-statistic
	Mean Velocity (MV)	Peak Velocity (PV)	Mean Velocity (MV)	Peak Velocity (PV)
40%	0.92	0.50	1073.66	97.15
50%	0.88	0.26	723.49	34.83
60%	0.76	0.51	301.48	101.32
All data	0.88	0.44	2086.1	234.72

*All data was significant with a p-value<0.001.

A linear regression model indicated a significant relationship between mean and peak bar velocity as reported by the GYM when compared to Qualisys tracking software. Mean velocity linear regression: F (1,293) = 2086.61, p<0.001, R2 = 0.88. Peak velocity linear regression: F (1,293) = 97.15, p < 0.001, R2 = 0.44. This model indicates that across all percentages of 1RM tested, GYM software was able to account for 88% of the variance in mean bar velocity and only 44% of peak bar velocity.
When parsed out and compared by loads, the data highlights a closer relationship between mean velocity measures as compared to peak velocity measures (Table 2.) At 40% 1RM: Mean velocity: F (1,293) = 1073.66, p < 0.001, R² = 0.92; Peak velocity: F (1,293) = 97.15, p < 0.001, R² = 0.50. At 50% 1RM: Mean velocity: F (1,293) = 723.49, p < 0.001, R² = 0.88; Peak velocity: F (1,293) = 34.83, p < 0.001, R² = 0.26. At 60% 1RM: Mean velocity: F (1,293) = 301.48, p < 0.001, R² = 0.76; Peak velocity: F (1,293) = 101.32, p < 0.001, R² = 0.51.
3.2 Reliability

Table 3. Intraclass Correlation Coefficients for mean and peak barbell velocity measurements.

	Mean Barbell Velocity	Peak Barbell Velocity
ICC (95% CI)	0.848 (0.341-0.941)	0.306 (-0.092-0.632)
F-statistic	23.6	4.8
p-value	0.00261	0.128

The ICCs were calculated to assess the reliability of mean and peak barbell velocity measurements. A two-way random-effects model with absolute agreement (ICC (A,1)) was used for both metrics. Mean barbell velocity had an ICC of 0.848 (0.341–0.941), with an associated F-test indicating statistical significance (F (296, 4.22) = 23.6, p = 0.00261). These calculations indicate good reliability. Peak barbell velocity had an ICC of 0.306 (-0.092–0.632), with a non-significant F-test (F (299, 2.69) = 4.8, p = 0.128). This ICC value indicates poor reliability.
The coefficients of variation (CV) were calculated to assess the relative variability in mean and peak values for both GYM and Qualisys datasets. For the mean values, the CV was 17.06% for GYM and 20.46% for Qualisys. For the peak values, the CV was 10.75% for GYM and 15.37% for Qualisys, with GYM showing the lowest relative variability among all measures.

Discussion
The findings of this study offer valuable insight into the reliability and validity of GYM as a VBT tool. While GYM demonstrated strong validity in tracking mean barbell velocity across all intensities, it was substantially less accurate when assessing peak barbell velocity. These results highlight important considerations for practitioners when using GYM as a training tool.
There was a strong correlation observed between GYM and Qualisys for mean velocity measurements, highlighting the reliability of GYM. The ICC for mean velocity (0.848) reflects good reliability, supporting its use by coaches and athletes where consistent data is essential for assessing training adaptations and adjusting programs accordingly. This finding demonstrates that GYM’s mean velocity measure is capable of providing practitioners with insightful data that can reliably indicate changes in athletes’ performance capabilities. For example, this means that a positive change of 0.15 m/s in an athletes mean clean velocity at a given load is likely due to changes in the athletes’ performance capabilities, as opposed to the measurement error associated with the VBT tool. This is rather important when competitive success has such slim margins and even more important when resistance training programs are dictated by real time data collected by VBT tools. These findings are consistent with prior research that has identified GYM as a reliable tool for monitoring barbell velocity during traditional resistance training exercises (17). Importantly, this examination focused on high velocity movements, hence the loads of 40-60%, and extended the range of velocities studied within the literature.
Despite this, GYM had a moderate correlation and systematically overestimated barbell velocity limiting its application. GYM had a mean bias of +0.37 m/s when assessing peak velocity suggesting that GYM may not offer the precision required for accurately evaluating peak velocity during rapid, explosive movements. What is perhaps more concerning is the poor ICC for peak velocity (0.306), indicating low reliability for this metric.. For example, if an athlete were to improve peak barbell velocity by 0.15 m/s, the same amount as with their mean velocity, we wouldn’t be able to confidently attribute this change to a performance improvement due to the low reliability.

These findings agree with previous research that has identified similar discrepancies in GYM’s accuracy. In Lorenzetti et al. (14), the GYM device showed a higher root mean square error (RMSE) of 0.06 m/s when assessing peak barbell velocity during ballistic jump squats compared to slower squat movements. This higher RMSE suggests that the device was less accurate in measuring peak velocity during higher velocity, explosive jumps. The study found the mean difference between GYM and the reference method (motion capture) to be -0.05 m/s, further indicating potential measurement errors in high-velocity movements. These results highlight that peak velocity measurements may be prone to greater variability in ballistic exercises. Additionally in Askow et al. (2), the GYM device consistently underestimated peak barbell velocities by 11.6% (or -0.13 m/s) when compared to a more accurate criterion measure. This bias was particularly evident during high-velocity movements, indicating that the device may not be as precise for measuring peak velocity in such contexts. The underestimation suggests a systematic error that could limit the utility of GYM for tracking performance improvements in peak velocity during explosive lifts. These values along with our data showcase that GYM may not be an effective tool at assessing peak barbell velocity at lower loads/higher barbell velocities.

This study also reinforces the importance of context when interpreting data from VBT devices. Contrary to our ICC data, the coefficients of variation (CV) highlight the consistency of GYM for both mean velocity (17.06%) and peak velocity (10.75%). Interestingly, this statistic suggests that peak velocity is more reliable when compared to mean velocity; however, this is likely due to the systematic overestimation of both peak and mean barbell velocity by GYM. Utilizing both ICC and CV’s the data supports the notion that GYM has strong reliability for mean velocity, however peak velocity measures capture by GYM leave something to be desired. These data suggest that practitioners should use mean barbell velocity measurements to achieve the best results, especially when utilizing VBT to monitor fatigue, track progress, and adjust training intensity in real time. Should practitioners have a penchant for peak velocity measures, the authors strongly encourage practitioners to run in-house statistics to understand what constitutes a meaningful change as compared to a change within the VBT’s measurement error.
Findings align with the broader literature discussing VBT devices and explore a gap in the literature by examining high-velocity movements while highlighting aspects that have practical significance. Future investigations should explore GYM’s performance with other high velocity movements such as the snatch or jerk, to better understand its broader applications. Importantly, while these results contribute to the growing body of evidence, it is important to situate the use of VBT within the broader training context and provide guidance to practitioners.

Application in Sport
The authors contend that reliable VBT tools can be leveraged by practitioners. First, VBT tools provide a cost-effective and time efficient avenue to collect data and highlight changes as a result of the training prescription. VBT data may be leveraged as biofeedback and a load modulation technique but only in synchrony with more traditional loading prescription (% of 1RM/% of set/rep best). Important to note, these strategies utilize VBT tools as a secondary data stream to inform when load changes may be needed and not as a primary load prescriber. Coaches must retain load prescription responsibilities, while utilizing their eyes and ears (in addition to VBT tools) to skillfully make load adjustments when needed. Practitioners must also bear in mind that VBT tools are inaccurate when estimating 1RM, therefore other methods for estimating are necessary. Perhaps the most compelling reason for utilizing VBT tools resides in their ability to potentiate participant performance. The presence of VBT devices may improve athlete motivation and training intent, which is paramount for optimal training. While VBT tools generally provide a positive return on investment, the practitioners’ eyes and ears should remain the primary data source which guide training decisions while VBT tools serve a supportive role. Based on available data, it would be shortsighted to rely solely on VBT tools to make real-time training decisions.

In conclusion, this study demonstrates that GYM provides reliable and valid measurements for mean barbell velocity during submaximal power cleans. As a result, practitioners may leverage GYM’s strengths, particularly its ability to provide immediate feedback and monitor mean velocity, while remaining cognizant of its limitations for high-velocity movements. This approach may allow for the effective integration of VBT tools to enhance training decisions, outcomes and athletic performance.

References:

Weakley JJS, Till K, Read DB, et al. The Effects of traditional, superset, and tri-set Resistance Training Structures on Perceived Intensity and Physiological Responses. European Journal of Applied Physiology. 2017;117(9):1877-1889. doi:https://doi.org/10.1007/s00421-017-3680-3

Weakley JJS, Wilson KM, Till K, et al. Visual Feedback Attenuates Mean Concentric Barbell Velocity Loss and Improves Motivation, Competitiveness, and Perceived Workload in Male Adolescent Athletes. Journal of Strength and Conditioning Research. 2019;33(9):2420-2425. doi:https://doi.org/10.1519/jsc.0000000000002133

Weakley J, McLaren S, Ramirez-Lopez C, et al. Application of velocity loss thresholds during free-weight resistance training: Responses and reproducibility of perceptual, metabolic, and neuromuscular outcomes. Journal of Sports Sciences. 2019;38(5):477-485. doi:https://doi.org/10.1080/02640414.2019.1706831

Weakley, J., Morrison, M., Garcia-Ramos, A., Johnston, R., James, L., and Cole, M. H. (2021). The validity and reliability of commercially available resistance training monitoring devices: A systematic review. Sports Medicine 51, 443-502.

Weakley J, Till K, Sampson J, et al. The Effects of Augmented Feedback on Sprint, Jump, and Strength Adaptations in Rugby Union Players Following a Four Week Training Programme. International Journal of Sports Physiology and Performance. 2019;14(9):1-21. doi:https://doi.org/10.1123/ijspp.2018-0523

Westcott WL. Resistance training is medicine: Effects of strength training on health. Current Sports Medicine Reports. 2012;11(4):209-216. doi:https://doi.org/10.1249/JSR.0b013e31825dabb8

Wilson KM, Helton WS, de Joux NR, Head JR, Weakley JJS. Real-time quantitative performance feedback during strength exercise improves motivation, competitiveness, mood, and performance. Proceedings of the Human Factors and Ergonomics Society Annual Meeting. 2017;61(1):1546-1550. doi:https://doi.org/10.1177/1541931213601750