You are on page 1of 29

Journal of Strength and Conditioning Research Publish Ahead of Print

DOI: 10.1519/JSC.0000000000002550

Using a Load-Velocity Relationship to Predict 1RM in Free-Weight Exercise: A

Comparison of the Different Methods

Hughes, Liam J. 1, Banyard, Harry G. 2, Dempsey, Alasdair R. 1, Scott, Brendan R. 1

D
TE
Author Information:
1
School of Psychology and Exercise Science, Murdoch University, Murdoch, Western

Australia
2
EP
Center for Exercise and Sports Science Research (CESSR), School of Medical and Health

Sciences, Edith Cowan University, Joondalup, Western Australia.

Address correspondence to Mr Liam Hughes:


C

Room 3.001, Social Sciences, 90 South Street, Murdoch WA 6150

Telephone: +61 4 00722618


C

Email : Liam.Hughes@murdoch.edu.au
A

No funding was received for this study.

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 1

Abstract

The purpose of this study was to investigate the reliability and validity of predicting 1-

repetition maximum (1RM) in trained individuals using a load-velocity relationship. Twenty

strength-trained men (age: 24.3±2.9 years, height: 180.1±5.9 cm, body mass: 84.2±10.5 kg)

were recruited and visited the laboratory on three occasions. The load-velocity relationship

D
was developed using the mean concentric velocity of repetitions performed at loads between

20% and 90% 1RM. Predicted 1RM was calculated using 3 different methods discussed in

TE
existing research; minimal velocity threshold 1RM (1RMMVT), load at zero velocity 1RM

(1RMLD0) and force-velocity 1RM methods (1RMFV). The reliability of 1RM predictions was

examined using intraclass correlation coefficient (ICC) and coefficient of variation (CV).
EP
1RMMVT demonstrated the highest reliability (ICC=0.92-0.96, CV=3.6-5.0%), followed by

1RMLD0 (ICC=0.78-0.82, CV=8.2-8.6%) and 1RMFV (ICC=-0.28-0.00, CV= N/A). Both

1RMMVT and 1RMLD0 were very strongly correlated with measured 1RM (r=0.91-0.95). The

only method which was not significantly different to measured 1RM was the 1RMLD0
C

method. However, when analyzed on an individual basis (using Bland-Altman plots), all

methods exhibited a high degree of variability. Overall, the results suggest that the 1RMMVT
C

and 1RMLD0 predicted 1RM values could be used to monitor strength progress in trained

individuals without the need for maximal testing. However, given the significant differences
A

between 1RMMVT and measured 1RM, and the high variability associated with individual

predictions performed using each method, they cannot be used interchangeably; therefore, it

is recommended that predicted 1RM is not used to prescribe training loads as has been

previously suggested.

Key Words: Squat, Maximal Strength, Resistance Training

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 2

INTRODUCTION

Assessments of maximal strength are commonly used to profile the physical capacities of an

individual (3, 35), prescribe loads to lift during training (24), and to track changes in

muscular abilities following a training intervention (2). While direct assessment of 1-

repetition maximum (1RM) to determine muscular strength is valid and reliable (33), these

assessments are time consuming, and physically and psychologically taxing, which can limit

D
their practicality (8, 16, 19). Numerous methods of estimating maximal strength have been

developed to predict 1RM from the number of repetitions performed with a given load (9, 25,

TE
34) or from multiple repetition maximum testing (31); however, despite requiring lighter

loads than 1RM tests, these methods are also time consuming and physically demanding.

Recent technological advances devices to quantify resistance exercise performance have


EP
resulted in additional methods to estimate maximal strength, using the inverse linear

relationship between the load lifted and the concentric velocity of a repetition (i.e. the load-

velocity relationship), which can be determined during a pre-training warm-up (7, 21, 22).
C

Early 1RM predictions from load-velocity relationships were based on the notion that the

final successful repetition of a set to failure will always occur at the same velocity (for the
C

same individual and exercise) (20); commonly referred to as the minimal velocity threshold

(MVT). The MVT-based 1RM prediction is performed by extrapolating the linear load-
A

velocity regression to its intersection with the MVT (4), to estimate the heaviest weight with

which an individual could complete a repetition at their specific MVT. Research has reported

strong correlations between the MVT-predicted 1RM (1RMMVT) and measured 1RM for

various Smith machine exercises (r = 0.93-0.95) (7, 21, 26). Multiple studies have also

suggested MVT is reliable for multi-joint exercises performed in a Smith machine (16, 20),

even with training-induced increases in actual 1RM (16). However, recent work from

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 3

Banyard et al. (4) has demonstrated poor test-retest reliability for the MVT variable when

using free-weight back squat exercise for well-trained subjects (coefficient of variation [CV]

= 22.5%). Furthermore, while 1RMMVT and measured 1RM were strongly correlated,

predicted 1RM scores significantly overestimated measured 1RM (4). The authors attributed

these differences to poor reliability of the MVT variable (4), suggesting that 1RMMVT

estimations may not be accurate enough to predict 1RM in the free-weight back squat

D
exercise.

TE
To overcome the potential limitations associated with 1RMMVT predictions, an alternative

method to estimate maximal strength involves extrapolating the load-velocity relationship to

the load which corresponds with a velocity of 0 m·s-1 (LD0), rather than the MVT (21). This

LD0 value will obviously be greater than 1RM; however, research indicates that the relative
EP
amount by which LD0 overestimates measured 1RM may be consistent for each individual

for a specific exercise (21). If the relative difference between LD0 and measured 1RM is

determined for each participant in key exercises, it can be subtracted from the LD0 value in
C

subsequent assessments to predicted 1RM (1RMLD0). However, previous research

investigating these 1RMLD0 predictions employed Smith machine based exercise, and the
C

reliability and validity of these predictions for free-weight exercise (which are associated

with more technical variance than machine-based exercise) is not known. It is also possible
A

that inter-individual differences in lifting kinematics (28, 29) will impact on the amount by

which 1RMLD0 overestimates 1RM, meaning that actual 1RM assessment would still be

required. In addition, the difference between LD0 and 1RM may be influenced by changes in

strength or alterations to lifting technique with training, and further investigation into this

predictive method is therefore warranted.

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 4

In light of recent interest in velocity-based 1RM predictions, a very recent study from Picerno

et al. (30) proposed another novel method to use velocity measures to predict 1RM. This

strategy examines the intersect between the regression line of the force-velocity relationship

(FV) with the weight (load x 9.81m·s-2)-velocity relationship (30). This point represents the

heaviest load at which the force an individual can produce is greater than the resisting weight

(i.e. predicted 1RM) (30). This force-velocity 1RM prediction (1RMFV) prediction was

D
almost perfectly related to measured 1RM for the pin loaded chest press and leg press

exercises

TE
(r > 0.99), and Bland-Altman analyses have demonstrated acceptable validity for the 1RMFV

using these exercises (chest press: bias = -1.32 kg, limits of agreement = -3.58-0.94 kg; leg

press: bias = -1.76 kg, limits of agreement = -5.81-2.29 kg). While this 1RMFV strategy
EP
appears very promising, research has not examined the validity of these predictions in free-

weight exercise, and the reliability of the predictions have not been published.

A potential limitation of current research is the emphasis of methodology employing


C

machine-based exercises. Importantly, the individuals who are most likely to undertake

velocity-based 1RM predictions are high-performance athletes training in a professional


C

setting; although these cohorts do make use of machine-based exercises, the majority of their

training is typically comprised of free-weight exercises (10). Therefore, the aim of this study
A

is to examine the reliability and validity of all currently available velocity-based 1RM

prediction methods using free-weight, rather than machine-based, exercise. It was

hypothesised that each of the load-velocity relationship based 1RM predictions would

demonstrate acceptable levels of test-retest reliability and validity, which would enable them

to be used in quantifying maximal strength.

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 5

METHODS

Experimental Approach to the Problem

Subjects reported to the laboratory on three occasions over 14 days, each separated by

approximately one week. The first visit served to familiarize subjects with the squat testing

protocol and the desired squatting technique, before determining squat 1RM and the minimal

velocity threshold (i.e. velocity of 1RM). This was followed by experimental Trial 1 and

D
Trial 2. During these trials subjects completed a warm-up protocol lifting progressively

heavier loads (20-90% 1RM), during which concentric velocity was monitored to establish

TE
individual load-velocity relationships. These data were subsequently used to predict 1RM

values based on MVT and LD0. Furthermore, these data were used in conjunction with force

data quantified in newtons to predict 1RM using the FV method. For Trial 1, subjects also
EP
performed another 1RM assessment following the warm-up protocol. This study design was

implemented to test the research hypothesis, specifically by assessing the test-retest reliability

of 1RM predictions between Trial 1 and Trial 2, and by examining the validity of these

predictions by comparing predicted and measured 1RM scores from Trial 1.


C

Subjects
C

Twenty male subjects (age: 24.3 ± 2.9 yr; height: 180.1 ± 5.9 cm; body mass: 84.2 ± 10.5 kg)

were recruited for this study. This sample size was calculated to detect a 7.5 kg difference
A

between actual and predicted 1RM scores with an effect size of 0.7, and was based on the

results of previous research (21). All subjects had ≥2 years resistance training experience,

were accustomed to performing the back squat exercise with correct technique, and exhibited

squat 1RM of ≥1.5 x body mass (measured squat 1RM = 151.1 ± 25.7 kg). Subjects were

excluded if there were currently injured, taking any performance enhancing supplements, or

had a health condition which could be exacerbated by exercise. During the study, subjects

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 6

were instructed not to perform any other strenuous physical activity. All subjects were

provided with information to detail the risks and benefits of participating, and signed an

informed consent form prior to beginning the experiment. The study and all its methodologies

were approved by the Institutional Human Ethics Committee.

Procedures

D
1-Repetition Maximum Testing

During the familiarization trial, 1RM was determined following established procedures (2, 6).

TE
Briefly, subjects completed a standardized warm-up comprising of five minutes on a cycle

ergometer at a moderate, self-selected intensity. This was followed by five minutes of

dynamic stretching, during which time subjects performed mobility exercises they typically
EP
employed prior to squat training. The specific mobility activities performed were recorded

and replicated during the warm-up for each subsequent trial. Subjects then completed warm-

up sets comprising of 3 repetitions at 20% of predicted 1RM (as estimated by the subject), 3

repetitions at 40% 1RM, 3 repetitions at 60% 1RM, 1 repetition at 80% 1RM and 1 repetition
C

at 90% 1RM (4), with 2 minutes of rest between each warm-up set. The weight was then

increased by ~5%, and subjects performed a single repetition. This process continued until
C

subjects were unable to successfully perform a lift with correct technique and given 3 minutes

of rest between attempts. The 1RM was defined as the heaviest completed repetition and was
A

determined within 3-6 sets.

During all trials, subjects were instructed to undertake the high-bar back squat exercise with

the eccentric phase performed under control, while the concentric phase was completed as

quickly as possible. With the barbell (Power Bar, Australian Barbell Company, Victoria,

Australia) supported on the superior trapezius, subjects flexed at the knees and hips until the

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 7

anterior aspect of the thighs were parallel with the ground. This bar was loaded with

International weightlifting federation approved weight plates (Ultimate Training Bumper

Plates, Force USA, Utah, USA). Subjects were instructed to position the front of their shoes

on a line marked on the floor beneath the squat rack, and to assume a stance width typical of

their normal squat technique. A customized elastic string-line was set to contact subjects’

superior hamstrings at this position to signal that appropriate depth had been reached (32),

D
which was visually confirmed by researchers positioned adjacent to the subject. Subjects

were also given verbal cues on when to halt the eccentric phase, and begin the concentric

TE
phase of the squat, along with consistent verbal encouragement (11).

Load-Velocity Profiling
EP
During Trial 1 and 2, subjects performed the same warm-up procedure as detailed for 1RM

testing, albeit using loads relative to the actual 1RM (rather than estimated 1RM). During

each lift, the displacement of the bar and time between data points were recorded by using a

cable linear position transducer sampling at up to 50Hz (23) (GymAware Powertool; Kinetic
C

Performance Technology, Canberra, Australia). From these data, the concentric phase of each

repetition was automatically identified by the linear position transducer, and the mean
C

concentric velocity was calculated for this portion of the lift. Furthermore, when the mass

lifted was incorporated into the linear position transducer software, this device also quantified
A

the force of each repetition in Newtons. This calculation is performed by first identifying the

acceleration of the system between data points (acceleration = change in velocity / change in

time), and then using this value to determine force (force = mass x acceleration) (13). The

specific linear position transducer used in this study has previously been validated as a means

of quantifying force in research which has compared it to direct measurement of force via a

force plate (13). The retractable cord of this device was fixed inside the barbell collar, with

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 8

the unit mounted on the floor directly beneath the bar’s position during the squatting action.

For warm-up sets that comprised of more than one repetition, the repetition with the fastest

mean concentric velocity was used for further analyses of the load-velocity profile as per

previous research (4, 21). Using the fastest repetition at each load ensures that the velocities

used to develop the load-velocity relationship represent the true best-performance of the

individual. With lighter loads (<80% 1RM) the fastest repetition is often likely to be the

D
second or third repetition, particularly in warm-up sets. This is possibly due to an increase in

the calcium flux from the sarcoplasmic reticulum, which improves the efficiency of muscle

TE
contractions (14). With heavier loads (≥80% 1RM) however, each additional repetition

performed is likely to induce fatigue, which would result in a decline in repetition velocity.

Post hoc analysis was undertaken to predict 1RM from these load-velocity profiles using
EP
three separate methods: 1) MVT method; 2) LD0 method, and; 3) FV method (Figure 1).

**INSERT FIGURE 1 NEAR HERE**


C

Data Analysis

The MVT for each individual was determined by measuring the velocity of the final
C

successful 1RM attempt from their initial familiarization trial. The 1RMMVT was calculated

by extrapolating each individual’s load-velocity regression line to their specific MVT, as per
A

Jovanović and Flanagan (22). Predicted 1RM for each subject was calculated primarily using

all of the load-velocity data points recorded (20-90% 1RM). As performing five warm-up sets

may not be practical in real-world setting, two additional four-load models were developed,

using 20-80% 1RM (i.e. excluding only the heaviest weight from the model) and 40-90%

1RM (i.e. excluding only the lightest weight from the model). As the MVT has been recently

shown to exhibit considerable variability (coefficient of variation [CV] = 22.5%) (4), we also

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 9

extrapolated the load-velocity regression line to its intercept with a velocity of 0 m·s-1 (LD0)

to provide another strength prediction as per Jidovtseff et al. (21). The percentage in which

the LD0 overestimated 1RM from the familiarisation session was then subtracted from this

value to determine the 1RMLD0 This 1RMLD0 analysis was also performed using the 3

different load combination described previously (20-90% 1RM, 20-80% 1RM and 40-80%

1RM). Finally, the 1RMFV was calculated using a custom Matlab (Math-Works; Natick,

D
Massachusetts) script with all calculations based on the supplementary material provided

Picerno et al. (30). This algorithm determines the intersection of the FV and the weight-

TE
velocity relationships (calculated by multiplying the loads used to develop the load-velocity

relationship by 9.81m·s-2). The load which corresponds with this intersection is the predicted

1RMFV (30). The 1RMFV predictions were also performed using the aforementioned load
EP
combinations (20-90% 1RM, 20-80% 1RM and 40-90% 1RM).

Statistical Analysis

The mean ± SD was calculated for all variables. The test-retest reliability of 1RMMVT,
C

1RMLD0 and 1RMFV predicted scores were determined by comparing 1RM predictions

between trials. The intraclass correlation coefficient (ICC: model 3, form 1) and coefficient
C

of variation (CV) with 95% confidence intervals were calculated using a custom made

spreadsheet designed for this purpose (18). This spreadsheet was also used to calculate the
A

ICC and CV values for individual points on the load-velocity relationship in Trial 1 and 2 to

examine reliability. Pearson’s product moment correlation coefficients were calculated to

assess the relationships between the measured and predicted 1RM scores. Bland-Altman plots

(5) were used to describe the level of agreement between measured and predicted 1RM

values. To identify any significant trends in these Bland-Altman plots, Pearson’s correlation

analysis was used to examine the relationship between the x (mean 1RM score) and y

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 10

(difference between predicted and measured 1RM) variables of each plot to identify any

trends in the data. A two-way ANOVA with repeated measures was performed to assess

differences in measured versus predicted 1RM scores during Trials 1 and 2, using a 4 x 2

design (1RM method x Trial). Where sphericity was violated Greenhouse-Geisser correction

procedure was used. In cases where a significant main effect was observed, Fisher’s Least

Significant Difference post hoc analyses were performed to determine where these

D
differences existed. These analyses assessed whether the 1RMMVT, 1RMLD0 and 1RMFV

predictions were reliable and could be used to accurately determine 1RM. Significance was

TE
set with a type-I error rate of α ≤ 0.05, and these analyses were performed using the SPSS

software (v.22, IBM, New York, USA).


EP
RESULTS

Mean back squat 1RM for the familiarization trial was 151.1 ± 25.7 kg. The mean measured

1RM for Trial 1 was 153.11 ± 26.8 kg and this variable demonstrated a high degree of test-

retest reliability between the two trials (ICC = 0.99, CV = 0.9%). Table 1 demonstrates the
C

test-retest reliability of repetitions performed at each of the individual velocities used to

develop the load-velocity relationship, as well as the velocity for 1RM repetitions. The most
C

reliable velocities were observed when lifting loads of 40-80% 1RM.


A

**INSERT TABLE 1 NEAR HERE**

Mean measured and predicted 1RM scores are presented in Table 2. There were no

significant differences for the 1RMMVT predictions between trials 1 and 2. For the 1RMMVT

there was no significant interaction effect between time and method of1RM assessment

method

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 11

(F2,35 = 0.715; p = 0.503; ƞ2 = 0.40). There was no significant main effect for time (F1,17 =

0.549; p = 0.469; ƞ2 = 0.031). There was a significant main effect for the method of 1RM

assessment (F2,31 = 17.431; p ≤ 0.01; ƞ2 = 0.506). Significant differences were observed for

the different strategies of 1RM assessment between measured 1RM and predicted 1RM using

20-90% 1RM (p < 0.001), 20-80% 1RM (p < 0.001) and 40-90% 1RM (p = 0.005). For the

1RMLD0 there was no significant interaction effect between time and method of 1RM

D
assessment method (F2,32 = 0.425, p = 0.425, ƞ2 = 0.024). There was no significant main effect

for time (F1,17 = 0.076; p = 0.786; ƞ2 = 0.004) nor was there significant main effect for the

TE
method of 1RM assessment (F2,29 = 0.674; p = 0.498; ƞ2 = 0.038). For the 1RMFV there was

no significant interaction effect (F1,17 = 0.965; p = 0.417; ƞ2 = 0.054). No significant main

effect was observed for time (F1, 17 = 1.020; p = 0.327; ƞ2 = 0.057), nor for method of 1RM

assessment (F1,17 = 0.976; p = 0.337; ƞ2 = 0.038).


EP
**INSERT TABLE 2 NEAR HERE**

The inter-session reliability of 1RM predictions is shown in Table 3. Slightly higher ICC and
C

lower CV values were associated with the MVT based 1RM prediction than the LD0.

However, ICC and CV values for all FV predictions were considerably poorer than all MVT
C

and LD0 predictive methods.


A

**INSERT TABLE 3 NEAR HERE**

All 1RM predictions from the MVT and LD0 methods exhibited very strong and significant

correlations with measured 1RM (r = 0.91-0.95). For the FV method, the only significant

correlation observed was for the 20-80% prediction in Trial 2 which demonstrated a

moderate-negative correlation (r = -0.50).

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 12

Figures 2-4 present Bland-Altman plots to describe the agreement between measured and

predicted 1RM models. The bias is indicated with a thick solid black line, the limits of

agreement with dotted black lines, and the relationship between x and y variables with a thin

black line. Significant relationships between x and y variables were observed for 1RMMVT

using loads of 20-80% 1RM (Figure 2b), and for all 1RMFV predictions (Figure 3) excluding

Trial 1 using loads of 20-80% 1RM.

D
These significant relationships identify that there is a threshold of strength which when

TE
surpassed leads to a greater magnitude of overestimation. For the 1RMMVT method performed

using loads of 20-80% 1RM the relationships suggest maximal strength is overestimated in

individuals with a 1RM greater than 106.1 kg for Trial 1 and 99.5 kg for Trial 2. For the
EP
1RMFV method this threshold is 151.90 kg in Trial 1, and 141.62 kg in Trial 2 for the 20-90%

1RM prediction, 153.53 kg in Trial 2 for the 20-80% 1RM prediction and 151.77 kg in Trial

1 and greater than 154.44 kg in Trial 2.


C

**INSERT FIGURES 2-4 NEAR HERE**


C

DISCUSSION

This study examined the reliability and validity of various 1RM prediction methods based on
A

the load-velocity relationship using the free-weight squat exercise. The main findings of this

investigation indicate; 1) 1RMMVT models exhibit high inter-session reliability and very

strong correlations with measured 1RM, although they seem to consistently overestimate

1RM, 2) 1RMLD0 also demonstrate high inter-session reliability and correlations between

predicted and measured 1RM, and 3) 1RMFV models display very poor inter-session

reliability and no significant positive relationship between predicted and measured 1RM.

Copyright ª 2018 National Strength and Conditioning Association

-
Using a Load-Velocity Relationship to Predict 1RM 13

While these pooled data indicate that 1RMMVT and 1RMLD0 methods might be acceptable for

predicting maximal strength, further analysis via Bland-Altman plots highlighted substantial

intra-individual variations between predicted and measured 1RM values.

The mean velocity during warm-up repetitions using 20% and 90% 1RM exhibited relatively

poor inter-test reliability (ICC = 0.25-0.38, CV = 7.2-14.2%). It is likely that reliability of

D
velocities measured with the 20% 1RM load were impacted by the fast velocities measured;

previous research has associated faster repetition speeds with a lower degree of limb

TE
coordination and more varied muscle activation patterns (1, 12, 27). Separately, the variance

in velocities at 90%, and also 100% 1RM, likely arose from technical differences between

repetitions performed with heavy weight. Indeed, Hay et al. (17) reported significant
EP
variations in the forward inclination of the trunk between repetitions of the squat exercise

with loads greater than 80% 1RM. The poor reliability of velocity at 1RM (ICC = 0.45, CV =

26.6%) observed in this study was similar to the results reported by Banyard et al. (4) (ICC =

0.42, CV = 22.5%) also for the free-weight squat. While these results suggest that load-
C

velocity profiling may be most reliable if weights ≤20% or ≥90% 1RM are excluded, we did

not observe this in our data (Table 2). A potential explanation for this is that the three
C

combinations of loads used to form load-velocity relationships in this study included at least

one of these ‘less reliable’ loads. If these predictions were made using loads between 40-80%
A

1RM, it is possible that reliability of 1RM predictions may be increased; however, this

approach was not used in our study as it would have limited the load-velocity profile to three

incremental loads (40%, 60% and 80% 1RM), and previous research has recommended that

at least four incremental loads are required to develop a load-velocity profile (22) .

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 14

Considering the 1RM predictive methods, the 1RMMVT method demonstrated the greatest

degree of reliability, closely followed by the 1RMLD0 method. Our reliability data indicate

that these 1RM predictions may be able to determine changes in strength between

assessments to within ~5% and ~9%, respectively. However, it must be acknowledged that

both of these predictive methods demonstrate poorer reliability when compared with direct

measurement of 1RM observed in this study (ICC = 0.99, CV = 0.9%). The usefulness of

D
these predictive methods may be limited for assessing maximal strength in response to a

training intervention, particularly for trained individuals who can only make small

TE
improvements in strength (36). In addition, the variability in these strength predictions limits

this method being used to estimate daily 1RM during a warm-up as a means of modulating

the loads to be lifted during training, as has been previously suggested by (22). Interestingly,
EP
the 1RM predictions performed using the 1RMFV method were shown to be unreliable for the

free-weight back squat exercise, which indicates that this method should not be used for free-

weight squat exercise.


C

Regarding the validity of velocity-based 1RM predictions, previous research investigating

1RMMVT methods has reported very strong correlations between the measured and estimated
C

1RM scores for the Smith machine bench press exercise (r = 0.95-0.98) (21). Our data

indicate similarly strong relationships between predicted and measured 1RM scores for both
A

1RMMVT (r = 0.91-0.95) and 1RMLD0 (r = 0.91-0.95) models. However, while previous

research has observed that 1RMFV is very strongly correlated with measured 1RM for the

machine chest press and leg press exercises (r = 0.99) (30), we did not observe similar

significant relationships in our study. It is likely that the mode of exercise is responsible for

these differences; Picerno et al. (30) utilized machine-based exercises, while the current study

employed a more technically demanding free-weight exercise. It became evident when

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 15

calculating the 1RMFV throughout this study that the predicted 1RM value was highly

sensitive to any change in each of the points used to develop both force-velocity and weight-

velocity relationships. 1RMFV calculations featuring faster than expected velocities with low

loads (20-40% 1RM) resulted in unrealistic overestimations of 1RM (e.g. 681.1 kg).

Conversely excessively slow velocities achieved with heavy loads (80-90% 1RM) seemed to

result in impossible underestimations of maximal strength (e.g. –62182.1 kg). Therefore, the

D
variance observed in the velocity of repetitions performed with 20% 1RM and 90% 1RM

(Table 1) is likely to have contributed to the differences in validity between the results of this

TE
paper and that published by Picerno et al. (30).

While the majority of research investigating the validity of velocity-based 1RM predictions
EP
has employed correlational analyses, it is important to acknowledge that this is often

inadequate and misleading when assessing agreement between two variables as they

represent the linear association and not the similarity of magnitudes (15). Indeed, recent work

from Banyard et al. (4) highlighted that despite strong correlations between 1RMMVT and
C

measured 1RM, predictive methods were associated with a large magnitude of error (SEE =

10.6-17.3kg,
C

CV = 7.4-12.8%). Our results concur with those from Banyard et al. (4), as both MVT and

LD0 predicted 1RM scores demonstrated a high degree of variability despite the very strong
A

relationships between measured and predicted values. To expand on this point further we

utilized Bland-Altman plots, which can account for within-subject variation and error in

assessing the relationship between a direct measure and an alternative technique (15). These

plots demonstrate substantial variability and poor agreement between measured and estimated

1RM scores across all predictive models. While the correlations and the analysis of the

pooled data seemed to suggest these methods may be effective to quantify maximal strength,

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 16

the Bland-Altman analysis demonstrated large overestimations and underestimations for

different subjects. In a practical setting, 1RM testing is performed to monitor individual

changes in maximal strength, and to prescribe individualized training load. Therefore, any

investigation into these methods of predicting 1RM should analyze validity at an individual

level, rather than in pooled statistical tests.

D
Significant trends were also observed in several of the Bland-Altman plots constructed in this

study, suggesting load-velocity 1RM predictions may overestimate strength in stronger

TE
individuals and underestimate strength in weaker individuals. It is important to consider

though that while these trends were statistically significant, they should be interpreted with

caution, as very large sample sizes are recommended to provide conclusive information on
EP
trends in Bland-Altman plots (5). While it is difficult to conduct very large laboratory-based

studies with sample sizes appropriate to investigate these trends, future research should aim

to determine whether there are differences between weaker and stronger individuals in the

usefulness of load-velocity 1RM predictions. Overall the results from this study highlight that
C

daily 1RM estimates based on load-velocity relationships may not be accurate enough to be

used for exercise prescription or to make adjustments to training load.


C

PRACTICAL APPLICATIONS
A

The results of this study demonstrate that the reliability of velocities achieved at very light

and very heavy loads is poorer than that using moderate loads. This indicates that load-

velocity profiles constructed with ≥ 4 incremental weights between 40-80% 1RM may be

most accurate. The 1RMMVT models provide the most reliable methods currently available to

predict 1RM from load-velocity profiles for free-weight squat exercise. However, these

models consistently over-estimated 1RM, meaning that predictive and measured 1RM values

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 17

cannot be used interchangeably (training load should not be prescribed as a percentage of

load-velocity relationship predicted 1RM values). While 1RMLD0 predictions were slightly

less reliable than 1RMMVT, the average overestimation of maximal strength was only ~4kg

(1RMLD0) instead of ~14 kg (1RMMVT). The 1RMLD0 method may therefore be considered for

use when the MVT for an individual or exercise is not known. Importantly though, Bland-

Altman analyses demonstrated substantial inter-individual variation in the accuracy of all

D
1RM predictions calculated, which may have been potentially masked by the basic

correlational analyses performed in previous studies. Therefore, our results emphasize that

TE
caution should be used when predicting 1RM values from load-velocity profiles, particularly

if it is the practitioner’s intention to use these values to adjust training load.

REFERENCES
EP
1. Almåsbakk B and Hoff J. Coordination, the determinant of velocity specificity? J

Appl Physiol 81: 2046-2052, 1996.

2. Baechle TR and Earle RW. Essentials of strength training and conditioning.


C

Champaign, IL: Human Kinetics, 2008.

3. Baker D and Nance S. The relation between running speed and measures of strength
C

and power in professional rugby league players. J Strength Cond Res 13: 230-235,

1999.
A

4. Banyard HG, Nosaka K, and Haff GG. Reliability and Validity of the Load-Velocity

Relationship to Predict the 1RM Back Squat. J Strength Cond Res, 2016.

5. Bland JM and Altman DG. Measuring agreement in method comparison studies. Stat

Methods Med Res 8: 135-160, 1999.

6. Bompa TO and Haff GG. Periodization: Theory and methodology of training.

Champaign, IL: Human Kinetics, 2009.

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 18

7. Bosquet L, Porta-Benache J, and Blais J. Validity of a commercial linear encoder to

estimate bench press 1 RM from the force-velocity relationship. J Sports Sci Med 9:

459-463, 2010.

8. Brechue WF and Mayhew JL. Upper-body work capacity and 1RM prediction are

unaltered by increasing muscular strength in college football players. J Strength Cond

Res 23: 2477-2486, 2009.

D
9. Brzycki M. Strength testing-predicting a one-rep max from reps-to-fatigue. J Phys

Educ Recreat Dance 64: 88-90, 1993.

TE
10. Cotterman ML, Darby LA, and Skelly WA. Comparison of muscle force production

using the Smith machine and free weights for bench press and squat exercises. J

Strength Cond Res 19: 169-176, 2005.


EP
11. Cronin JB and Hansen KT. Strength and power predictors of sports speed. J Strength

Cond Res 19: 349-357, 2005.

12. Darling WG and Cooke J. Movement related EMGs become more variable during

learning of fast accurate movements. J Motor Behav 19: 311-331, 1987.


C

13. Dorrell H, Moore J, Smith MF, and Gee T. Validity and whole system reliability of a

commercially available linear positional transducer across common resistance training


C

exercises. National Strength and Conditioning Association, 2017.

14. Endo M. Calcium release from the sarcoplasmic reticulum. Physiological Reviews 57:
A

71-108, 1977.

15. Giavarina D. Understanding Bland-Altman analysis. Biochem Med 25: 141-151,

2015.

16. González-Badillo JJ and Sánchez-Medina L. Movement velocity as a measure of

loading intensity in resistance training. Int J Sports Med 31: 347-352, 2010.

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 19

17. Hay JG, Andrews JG, Vaughan CL, and Ueya K. Load, speed and equipment effects

in strength-training exercises, in: Biomechanics VII-B. Champaign, IL: Human

Kinetics, 1983, pp 939-950.

18. Hopkins W. A new view of statistics. 2002. Available at:

http://www.sportsci.org/resource/stats/xrely.xls. Accessed July 1, 2017.

19. Horvat M, Franklin C, and Born D. Predicting strength in high school women

D
athletes. J Strength Cond Res 21: 1018-1022, 2007.

20. Izquierdo M, González-Badillo J, Häkkinen K, Ibanez J, Kraemer W, Altadill A,

TE
Eslava J, and Gorostiaga EM. Effect of loading on unintentional lifting velocity

declines during single sets of repetitions to failure during upper and lower extremity

muscle actions. Int J Sports Med 27: 718-724, 2006.


EP
21. Jidovtseff B, Harris NK, Crielaard J-M, and Cronin JB. Using the load-velocity

relationship for 1RM prediction. J Strength Cond Res 25: 267-270, 2011.

22. Jovanović M and Flanagan EP. Researched applications of velocity based strength

training. J Aust Strength Cond 22: 58-69, 2014.


C

23. Kinetic Perfomance. GymAware Sampling Method. Available at:

https://kinetic.com.au/pdf/sample.pdf. Accessed December 20, 2017.


C

24. Kraemer WJ and Ratamess NA. Fundamentals of resistance training: progression and

exercise prescription. Med Sci Sports and Exerc 36: 674-688, 2004.
A

25. Kravitz L, Akalan C, Nowicki K, and Kinzey S. Prediction of 1 repetition maximum

in high-school power lifters. J Strength Cond Res 17: 167-172, 2003.

26. Loturco I, Kobal R, Moraes JE, Kitamura K, Abad CCC, Pereira LA, and Nakamura

FY. Predicting the maximum dynamic strength in bench press: The high precision of

the bar velocity approach. J Strength Cond Res 31: 1127-1131, 2017.

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 20

27. Marsden C, Obeso J, and Rothwell J. The function of the antagonist muscle during

fast limb movements in man. J Physiol 335: 1-13, 1983.

28. McKean MR, Dunn PK, and Burkett BJ. Quantifying the movement and the influence

of load in the back squat exercise. J Strength Cond Res 24: 1671-1679, 2010.

29. Newton RU, Murphy AJ, Humphries BJ, Wilson GJ, Kraemer WJ, and Häkkinen K.

Influence of load and stretch shortening cycle on the kinematics, kinetics and muscle

D
activation that occurs during explosive upper-body movements. Eur J Appl Physiol

Occup Physiol 75: 333-342, 1997.

TE
30. Picerno P, Iannetta D, Comotto S, Donati M, Pecoraro F, Zok M, Tollis G, Figura M,

Varalda C, and Di Muzio D. 1RM prediction: a novel methodology based on the

force–velocity and load–velocity relationships. Eur J Appl Physiol 116: 2035-2043,


EP
2016.

31. Reynolds JM, Gordon TJ, and Robergs RA. Prediction of one repetition maximum

strength from multiple repetition maximum testing and anthropometry. J Strength

Cond Res 20: 584-592, 2006.


C

32. Scott BR, Dascombe BJ, Delaney JA, Elsworthy N, Lockie RG, Sculley DV, and

Slattery KM. The validity and reliability of a customized rigid supportive harness
C

during Smith Machine back squat exercise. J Strength Cond Res 28: 636-642, 2014.

33. Seo D-i, Kim E, Fahs CA, Rossow L, Young K, Ferguson SL, Thiebaud R, Sherk VD,
A

Loenneke JP, and Kim D. Reliability of the one-repetition maximum test based on

muscle group and gender. J Sports Sci Med 11: 221, 2012.

34. Shimano T, Kraemer WJ, Spiering BA, Volek JS, Hatfield DL, Silvestre R, Vingren

JL, Fragala MS, Maresh CM, and Fleck SJ. Relationship between the number of

repetitions and selected percentages of one repetition maximum in free weight

exercises in trained and untrained men. J Strength Cond Res 20: 819-823, 2006.

Copyright ª 2018 National Strength and Conditioning Association


Using a Load-Velocity Relationship to Predict 1RM 21

35. Wisløff U, Castagna C, Helgerud J, Jones R, and Hoff J. Strong correlation of

maximal squat strength with sprint performance and vertical jump height in elite

soccer players. Br J Sports Med 38: 285-288, 2004.

36. Zatsiorsky VM and Kraemer WJ. Science and practice of strength training.

Champaign, IL: Human Kinetics, 2006.

D
LIST OF FIGURES

Figure 1. Examples of load-velocity relationships used to predict 1RM (predicted 1RM value

TE
indicated by ); A) MVT method with load-velocity relationship (dotted black line), MVT
(grey line) and predicted 1RM score ( ) labelled. B) LD0 method with load-velocity
relationship (dotted black line), LD0 overestimation (grey line) and predicted 1RM score ( )
labelled. C) FV method with force- velocity relationship (dotted black line), weight-velocity
relationship (solid grey line) and predicted 1RM score ( ) labelled.
Figure 2. Bland-Altman plots representing the difference in measured 1RM and the
EP
1RMMVT calculated using loads of (a) 20-90% 1RM, (b) 20-80% 1RM and (c) 40-90%
1RM against the mean 1RM values across both methods. The horizontal solid line represents
the bias and the horizontal dotted lines represent the limits of agreement. The relationship
between x and y variables is represented by the sloping black line, with the Pearson’s product
moment correlation and its significance noted in the bottom left of each plot.
Figure 3. Bland-Altman plots representing the difference in measured 1RM and the 1RMLD0
C

calculated using loads of (a) 20-90% 1RM, (b) 20-80% 1RM and (c) 40-90% 1RM against
the mean 1RM values across both methods. The horizontal solid line represents the bias and
the horizontal dotted lines represent the limits of agreement. The relationship between x and
C

y variables is represented by the sloping black line, with the Pearson’s product moment
correlation and its significance noted in the bottom left of each plot.

Figure 4. Bland-Altman plots representing the difference in measured 1RM and the 1RMFV
A

calculated using loads of (a) 20-90% 1RM, (b) 20-80% 1RM and (c) 40-90% 1RM against
the mean 1RM values across both methods. The horizontal solid line represents the bias and
the horizontal dotted lines represent the limits of agreement. The relationship between x and
y variables is represented by the sloping black line, with the Pearson’s product moment
correlation and its significance noted in the bottom left of each plot. It is important to note
that the axis values on each plot are different and must be taken into consideration when
interpreting this figure. Furthermore, a data point (-31021.23, 62322.46) has been excluded
from (a) Trial 2 so as to not distort the graphic representation of the other data.

Copyright ª 2018 National Strength and Conditioning Association


Table 1: Change in mean, intraclass correlation coefficients and coefficients of variation for
test-retest assessments of the velocity of different loads.

∆ ± SD ICC (CI) CV (CI)

Mean Velocity 20%1RM 0.06 ± 0.07 0.38 (-0.17-0.68) 7.2% (5.4-11.5)

Mean Velocity 40%1RM 0.05 ± 0.04 0.74 (0.39-0.87) 6.4% (4.9-9.3)

Mean Velocity 60%1RM 0.04 ± 0.03 0.62 (0.15-0.81) 2.5% (2.0-3.7)

Mean Velocity 80%1RM 0.02 ± 0.3 0.54 (0.00-0.77) 4.3% (3.3-6.3)

D
Mean Velocity 90%1RM 0.04 ± 0.06 0.25 (-0.29-0.63) 14.2% (10.5-22.9)

Mean Velocity 100%1RM 0.03 ± 0.03 0.45 (0.32-0.56) 22.6% (19.3-26.0)

TE
∆ ̅ = change in mean velocity, ICC = intraclass correlation coefficient, CV = coefficient of variation, CI = 95%
confidence intervals.
Note: The test-retest reliability of the mean velocity at 100% 1RM was examined between the familiarization
session and Trial 1 as this load was not used in Trial 2.
EP
C
C
A

Copyright ª 2018 National Strength and Conditioning Association


Table 2: Mean measured and predicted 1RM values in a Trial 1 and Trial 2

1RM Assessment Method Trial 1 Trial 2

Measured 1RM (kg) 153.11 ± 26.8 -

Predicted 1RMMVT (kg) 20-90% 163.4 ± 31.2* 165.0 ± 32.5*

20-80% 166.8 ± 34.3* 170.11 ± 34.9*

40-90% 159.8 ± 29.6* 160.11 ± 30.57*

Predicted 1RMLD0 (kg) 20-90% 153.6 ± 29.0 152.23 ± 29.9

D
20-80% 152.9 ± 30.7 153.75 ± 29.2

40-90% 156.0 ± 29.4 152.92 ± 30.4

TE
Predicted 1RMFV (kg) 20-90% 158.7 ± 78.3 -3267.5 ± 14704.1

20-80% 155.2 ± 35.5 159.5 ± 81.4

40-90% 197.7 ± 176.4 126.1 ± 102.4


EP
1RM = 1-repetition maximum, 1RMMVT = 1RM predicted via minimal velocity threshold method, 1RMLD0 =
1RM predicted via load at zero velocity method and 1RMFV = 1RM predicted via force-velocity method.
*Significantly different to measured 1RM.
C
C
A

Copyright ª 2018 National Strength and Conditioning Association


Table 3: Change in mean, intraclass correlation coefficient and coefficient of variation for
test, retest assessments of 1RM predicted from the load-velocity relationship.

Predictive Method ∆ ± SD ICC (CI) CV (CI)

1RMMVT 20-90% (kg) 2.17 ± 1.24 0.92 (0.82-0.97) 5.0% (3.9-7)

20-80% (kg) 3.33 ± 0.63 0.92 (0.84-0.97) 4.9% (3.9-7)

40-90% (kg) 0.28 ± 0.96 0.96 (0.91-0.98) 3.6% (2.8-5.1)

1RMLD0 20-90% (kg) 1.40 ± 0.87 0.82 (0.63-0.90) 8.2% (6.4-11.6)

D
20-80% (kg) 0.89 ± 1.45 0.78 (0.57-0.9) 8.5% (6.6-12.1)

40-90% (kg) 3.08 ± 0.99 0.81 (0.62-0.91) 8.6% (6.7-12.3)

TE
1RMFV 20-90% (kg) 3426.20 ± 14675.11 0.00 (-0.39-0.39) -

20-80% (kg) 4.38 ± 45.81 -0.28 (-0.6-0.12) -

40-90% (kg) 29.85 ± 74.06 -0.11 (-0.48-0.29) -

∆ ̅ = change in mean velocity, ICC = intraclass correlation coefficient, CV = coefficient of variation, CI = 95%
EP
confidence intervals.
Note: Coefficient of variation (CV) can only be calculated for ratio data, due to the 1RMFV prediction method
resulting in negative values the CV for this method could not be calculated.
C
C
A

Copyright ª 2018 National Strength and Conditioning Association


D
TE
EP
C
C
A

Copyright ª 2018 National Strength and Conditioning Association


D
TE
EP
C
C
A

Copyright ª 2018 National Strength and Conditioning Association


D
TE
EP
C
C
A

Copyright ª 2018 National Strength and Conditioning Association


D
TE
EP
C
C
A

Copyright ª 2018 National Strength and Conditioning Association