Load-Velocity Profiling for 1RM Prediction: Accuracy Review

The one-repetition maximum (1RM) is the gold standard measure of maximal strength capacity and the anchor for load prescription in percentage-based training. However, direct 1RM testing involves true maximal effort, carries musculoskeletal injury risk (estimated at 1–3 per 1,000 testing sessions in trained athletes), and requires significant time when managing a squad of athletes across multiple exercises.

Load-velocity profiling (LVP) — establishing the relationship between barbell load and mean concentric velocity across submaximal loads — emerged from Gonzalez-Badillo and Sanchez-Medina's seminal 2010 observation that the load-velocity relationship in the squat is highly linear (R² ≥ 0.97) and that the 1RM load corresponds to a predictable minimum velocity for each individual. If a practitioner can measure velocity at 3–5 submaximal loads and establish this linear relationship, they can extrapolate to the load corresponding to the individual's minimum velocity threshold — yielding an estimated 1RM without a maximal effort attempt.

The key question this review addresses is: how accurate is this prediction in practice, and what factors most affect its precision?

A standard LVP session proceeds as follows:

1. The athlete performs sets at progressively increasing loads (e.g., 40%, 55%, 70%, 85% of estimated 1RM), performing 3 repetitions at each load with full concentric effort and maximal velocity intent.
2. Mean concentric velocity (MCV) is recorded at each load using a linear position transducer or IMU sensor.
3. A linear regression is fitted to the load-MCV data: Load = a − (b × MCV), where a and b are individually determined slope and intercept parameters.
4. The 1RM is estimated by inserting the individual's minimum velocity threshold (MVT — the slowest MCV at which the athlete can complete a repetition) into the regression equation: estimated 1RM = a − (b × MVT).

The accuracy of the estimate depends on: (1) the quality of velocity measurement at each load, (2) the selection of loads spanning a wide enough velocity range, (3) the accuracy of the individual MVT, and (4) how recently the profile was established relative to the current training state.

The squat is the most-studied exercise for LVP-based 1RM prediction, with consistent evidence across multiple laboratories and populations:

• Gonzalez-Badillo and Sanchez-Medina (2010) — the landmark study. Using 4–6 submaximal loads (40–93% 1RM) in 80 trained men, the load-velocity relationship explained 97.2–99.5% of variance (R² across participants). The standard error of estimate (SEE) for 1RM prediction using extrapolation to a population-average MVT of 0.31 m/s was 3.2 kg.
• Garcia-Ramos et al. (2018) — validated LVP in 30 powerlifters. Using individually calibrated MVT, SEE = 2.9 kg (3.1% of 1RM). Using population-average MVT, SEE increased to 5.4 kg (5.8%) — demonstrating the importance of individual MVT calibration.
• Weakley et al. (2020) — systematic review of squat LVP accuracy. Across 12 studies (n = 437), mean SEE = 3.8 kg (range: 2.1–6.4 kg). Greater accuracy was associated with: higher sampling rate (≥800 Hz), more loads in the profile (4+ loads), wider load range (≥40% 1RM span), and individually calibrated MVT.

Practically, a well-executed squat LVP estimates 1RM within ~4 kg in a trained lifter — which is sufficient for percentage-based load prescription, where ±5% error in 1RM translates to ±2.5–5 kg variation in prescribed loads.

Bench press LVP accuracy is systematically lower than squat accuracy, driven by two factors: greater inter-individual MVT variation and the greater influence of bar path inconsistency on velocity measurement.

• Garcia-Ramos et al. (2016) compared LVP 1RM prediction vs directly tested 1RM in 30 trained men (bench press). Using population-average MVT (0.17 m/s), SEE = 7.1 kg (6.8% of mean 1RM of 105 kg). Using individually calibrated MVT: SEE = 4.4 kg (4.2%) — a 38% improvement in accuracy.
• Randell et al. (2021) — 40 resistance-trained subjects, bench press LVP using 4 loads. Mean absolute error = 5.9 kg (range: 2.1–12.8 kg). The wide individual range highlights that some athletes deviate substantially from the population MVT assumption, making individual calibration critical.
• Bar path deviation in the bench press (medial-lateral movement, varying arc trajectory) introduces additional velocity measurement error compared to the more constrained squat bar path. Studies using a linear position transducer attached to the bar show lower bench press SEE than IMU-based methods, likely due to this constraint (Garcia-Ramos et al., 2016).

For bench press, the recommendation is to use individually calibrated MVT rather than population averages. Population-average MVT values for the bench press range from 0.14 to 0.22 m/s (SD ±0.06 m/s) — a range large enough to cause errors of 8–15 kg if the individual deviates significantly from the mean.

A pragmatic concern for coaches is the number of loads required for a valid LVP. A full 4–6-load protocol requires a dedicated testing session, while a two-point protocol (using only a light and a heavy load) could be integrated into a normal warm-up sequence.

The two-point LVP concept was formalized by Jaric et al. (2015) and validated specifically for strength testing by Garcia-Ramos et al. (2018) and others. The key evidence:

• Garcia-Ramos et al. (2018) compared 2-load (45% + 85% 1RM) vs 4-load (45%, 60%, 75%, 85% 1RM) LVP for squat 1RM prediction in 26 trained men. 2-load SEE = 3.4 kg vs 4-load SEE = 2.9 kg — a non-significant difference (p = 0.31). Bland-Altman analysis showed comparable limits of agreement.
• Critical condition: the two loads must be separated by at least 30% 1RM (preferably ≥35–40%). When separation was < 20%, 2-load LVP accuracy deteriorated significantly (SEE increasing to 6.8–9.2 kg), as the regression extrapolates over a wider range from a poorly constrained line.
• Practicality advantage: a 2-load LVP at 45% and 85% takes approximately 8–12 minutes including inter-set rest, compared to 20–35 minutes for a full 4–6-load protocol. This makes it feasible to re-profile athletes weekly in time-constrained environments.

The two-point method is now the preferred approach in most high-performance environments for routine monitoring. The full multi-load profile is reserved for initial individual MVT calibration and post-macrocycle assessments.

The minimum velocity threshold (MVT) is the most consequential source of LVP prediction error. If the MVT used in the calculation deviates from the individual's true MVT, the estimated 1RM shifts by approximately 8–15 kg per 0.05 m/s of MVT error — a direct and large effect on prediction accuracy.

Published MVT values from large-sample studies:

Data from Gonzalez-Badillo and Sanchez-Medina (2010), Garcia-Ramos et al. (2016), Weakley et al. (2020).

Individual MVT determination requires a genuine 1RM attempt at least once — the velocity during this attempt is recorded as the individual's MVT. Thereafter, all LVP sessions use this individual calibration rather than population averages. MVT itself is relatively stable: Jukic et al. (2020) showed test-retest reliability of individual MVT across 4 weeks was ICC = 0.94, suggesting MVT does not need to be re-determined at every training block — but should be re-established when technique changes significantly or after a long detraining period.

A critical but under-studied aspect of LVP practice is how long a profile established on one date remains an accurate predictor of current 1RM. As strength improves over a training block, the load-velocity relationship shifts — the athlete moves faster at the same absolute load, meaning the old regression line no longer accurately extrapolates to the current 1RM.

• Weakley et al. (2020) found that profiles established at the start of an 8-week training block predicted 1RM at the end of the block with SEE of 7.2 kg — more than double the accuracy of a contemporaneous profile (3.8 kg). The deterioration in accuracy occurred primarily in athletes who gained significant strength (> 5% increase in 1RM), where the shifted load-velocity relationship was no longer captured by the old profile.
• Jukic et al. (2020) demonstrated that 1RM prediction error from an 8-week-old profile was clinically meaningful: using stale profiles resulted in load prescriptions that were off by 6–12% of 1RM — the equivalent of prescribing 70% effort when 62% was the actual target. For VBT or percentage-based training, this is a functionally significant error.
• Practical recommendation: re-establish the LVP every 4–6 weeks during active training, or whenever a ≥5% change in 1RM is suspected. The two-point method makes this practical without requiring a dedicated testing session.

Based on the evidence reviewed, the following LVP protocol is recommended for training environments:

Initial Calibration (perform once, repeat every macrocycle)

1. Perform a full 4–6-load LVP spanning 40–90% estimated 1RM with 3 reps per load (maximal intent, minimal velocity loss).
2. At the end of the session, perform a genuine 1RM attempt and record the velocity — this is the individual MVT for the exercise.
3. Store this MVT value; use it for all subsequent routine monitoring sessions.

Routine Monitoring (every 2–4 weeks)

1. As part of a normal warm-up, perform 3 reps at ~45% 1RM and 3 reps at ~80–85% 1RM with full velocity intent.
2. Fit the 2-point regression using the individual MVT to estimate current 1RM.
3. Compare estimated 1RM to previous value — if ≥5% change, adjust load prescriptions accordingly.

Key practical standards

• Always use maximal concentric intent on every rep, regardless of load — submaximal effort inflates measured velocity at light loads, distorting the regression.
• Use 3 reps per load; discard the first rep (warm-up effect) and average the remaining two for the velocity data point.
• Minimum load separation for the two-point method: 35% 1RM between the two loads.
• Re-establish individual MVT whenever technique changes substantially or after ≥6 weeks of detraining.