Load-Velocity Profiling: Can Bar Speed Predict Your 1RM?

Ask a velocity-based training app what you can squat today, and it will hand you a number in kilograms within seconds. No warm-up ramp to a grinding single, no guessing whether last week's poor sleep ate into your ceiling. The bar moves at a certain speed, the app runs that speed through a load-velocity equation, and a one-repetition maximum appears on screen. It looks like measurement. The evidence says it is closer to a well-informed estimate, and the size of the error matters more than most lifters realise.

Load-velocity profiling sits at the centre of velocity-based training (VBT), the practice of using bar speed rather than a fixed percentage of 1RM to set training load. The appeal is obvious: strength fluctuates day to day with sleep, stress, and prior training load, and a percentage-based programme written six weeks in advance cannot account for any of that. If velocity reliably tracks proximity to true maximum strength, a wearable that reads bar speed in real time should outperform a spreadsheet written in isolation. That premise has driven a decade of linear position transducers, barbell-mounted sensors, and smartphone apps, all promising to replace the 1RM test. The research on whether that promise holds is now mature enough to give a clear, if less flattering, answer.

What Load-Velocity Profiling Actually Claims

The method rests on a simple physical relationship: as load increases toward 1RM, concentric velocity decreases in a predictable, close to linear pattern for a given exercise and lifter.¹ Plot velocity against load across a handful of submaximal reps and you can extrapolate the line to the point where velocity would theoretically reach zero, the minimal velocity threshold, and read the corresponding load as an estimated 1RM. Two-point and multi-point testing protocols exist, alongside generalized equations built from population data so a lifter never has to test at all.² The relationship has been demonstrated across the back squat, bench press, deadlift, and leg press, and it underpins most commercial VBT software: log a few warm-up sets, and the app returns a projected maximum and training zones for strength, hypertrophy, and power without a true single ever being attempted.

How Accurate Is the Estimate, Really

The largest test of this claim is a 2023 individual participant data meta-analysis pooling 137 prediction models across 26 studies.³ Across that dataset, load-velocity predictions carried a standard error of estimate around 9.8%, and a consistent bias toward overestimating true 1RM by roughly 4.5 kg, about 3.7% above the measured value. That is a moderate level of validity at the group level, useful for tracking a trend, not precise enough to programme a competition attempt off a single reading.

Part of the problem sits inside the equation itself. The foundational reliability study on the back squat found that while an actual 1RM was extremely stable across repeated trials (ICC 0.99), the velocity recorded at that same 1RM was not: it varied by up to 22.5% between sessions, with a between-trial reliability of only 0.42.¹ The authors concluded plainly that the method "cannot accurately modify sessional training loads" because the anchor point it depends on, the velocity at true failure, moves around too much to trust. A follow-up study on deadlift and back squat 1RM estimation found the same pattern from a different angle: at lighter submaximal loads (40 to 80% of 1RM), session-to-session reliability was poor (ICC as low as 0.17), and only improved to a strong, dependable range once testing moved closer to true maximal loads.⁴ In other words, the lighter and safer the warm-up sets a lifter is willing to use for profiling, the less trustworthy the resulting estimate becomes.

Age adds a further wrinkle. A 2026 study in adults aged 55 to 81 found that raw velocity-based estimates on a leg press systematically overestimated true 1RM, and that even after applying a mathematical correction for the non-linear relationship near maximal loads, group-level accuracy improved but meaningful interindividual variability persisted.⁵ The correction helped the average lifter in the sample. It did not guarantee the number was right for any one of them.

Generalized Equations vs Personal Testing: A Surprising Null Result

Given how much individual variation shows up in the data above, the intuitive fix is to build a personalized load-velocity profile for each lifter rather than relying on a population-derived equation. The evidence does not support that intuition. The same 2023 meta-analysis found only limited differences in predictive accuracy between individualized and generalized load-velocity relationships.³ A 2026 bench press study went further, directly comparing generalized against individualized equations for setting relative load in twenty trained men and finding no significant differences in repetitions completed, velocity loss patterns, or post-set fatigue between the two approaches.⁶ Spending a session establishing your own personal velocity curve, in other words, buys less certainty than it feels like it should. The noise sits in the biology of the velocity signal itself, not primarily in whose data built the formula.

If the Number Is Imprecise, Does VBT Still Work

None of this means velocity-based training is a wasted approach, and here the literature separates two different questions that get conflated in most marketing copy: does velocity accurately predict a single absolute number, and does training by velocity produce good outcomes. A systematic review and meta-analysis of load and volume autoregulation found no significant difference in strength gains between velocity-based and standardized percentage-based programming, meaning VBT is not inferior to a well-built spreadsheet, but it is not shown to be clearly superior either.⁷ What did matter inside VBT protocols was the velocity-loss threshold chosen within a set: lower thresholds preserved velocity and favoured strength, while higher thresholds accumulated more volume and favoured hypertrophy.⁷ A separate meta-analysis in trained athletes found VBT produced meaningful improvements over comparison training in maximum strength and countermovement jump performance.⁸ Both findings describe the value of using velocity as a real-time regulator of effort and volume within a session, not the value of the extrapolated 1RM number itself.

There is a further, exercise-specific caveat worth flagging. Even the minimal velocity threshold at 1RM is not a fixed constant; it needs to be established per exercise, and studies on movements like the hexagonal barbell deadlift have had to determine the optimal threshold specifically for that lift rather than borrowing one from the squat or bench press.⁹ A 2026 analysis of load-velocity profiles across weightlifting exercises reinforced this, finding that the popular idea of universal velocity zones (roughly 1.0 to 0.75 m/s for "starting strength", for example) does not consistently hold across different movements, meaning a training zone calibrated on the squat may be meaningless applied to the snatch or clean pull.¹⁰

The Measurement Layer Nobody Questions

Every figure above assumes the device recording bar speed is itself accurate, and that assumption deserves its own scrutiny. Validation work comparing a smartphone AI application against a criterion linear position transducer during the bench press found a very high correlation at both 50% and 75% of 1RM (r of 0.90 and 0.92), with the app's own test-retest coefficient of variation running 8.2% to 13.6%, broadly comparable to the criterion device's 6.5% to 12.1%.¹¹ That is a genuinely good result for a consumer tool, and it also illustrates the ceiling: even a validated device carries roughly 8 to 13% of its own measurement noise, which then stacks on top of the 9 to 10% biological noise in the load-velocity relationship itself, and the 20%-plus variability in velocity at true 1RM described above. None of these error sources are enormous individually. Compounded, they explain why a single velocity reading translated into a precise kilogram figure should be treated as a directional estimate rather than a certified number, regardless of which device produced it.

What This Means in Practice

The honest takeaway is not that velocity is a poor training signal. It is that velocity answers a narrower question than the marketing around it suggests. It is a strong within-session and within-week signal: is this rep meaningfully slower than the last one, is today's bar speed at a given weight trending down against last month's average, has the drop-off across a set crossed the threshold you decided in advance was your stopping point. Those relative comparisons are where the reliability data above holds up best, because they cancel out much of the day-to-day noise in the absolute number. What velocity is a weaker tool for is standing in for a true 1RM test on a single reading, particularly from light warm-up sets.

This is why ZELOS treats the IMU stream from Z1 as a relative signal first. Bar and limb velocity during a set are shown as trend, comparing this rep to your recent baseline and this session to your history, rather than asserted as a precise absolute load figure the wearable cannot fully defend. It sits alongside the same caution already built into ZELOS's approach to stopping a set on velocity loss: useful for the decision of when to ease off within a working set, not a substitute for occasionally testing a true heavy single to keep the whole system anchored to reality.

Key Takeaways

• Load-velocity equations predict 1RM with roughly 9 to 10% error at the group level and a systematic overestimate of about 3 to 4%, which is useful for tracking a trend but not precise enough for a single high-stakes decision.

• Velocity recorded at true 1RM is itself unstable between sessions, varying more than 20% in some studies, which caps how accurate any downstream equation can be.

• Personalized load-velocity profiles are not meaningfully more accurate than generalized population equations, so building your own curve buys less certainty than expected.

• The training benefits credited to velocity-based training come from real-time autoregulation of effort and volume within a session, not from the precision of an extrapolated 1RM number.

• Even validated velocity-measurement devices carry their own 8 to 13% measurement noise, which compounds with the biological noise above, so any single absolute reading deserves a healthy margin of doubt.

References

1. Banyard HG, Nosaka K, Haff GG. Reliability and validity of the load-velocity relationship to predict the 1RM back squat. J Strength Cond Res. 2017;31(7):1897-1904.

2. Ramos AG. Resistance training intensity prescription methods based on lifting velocity monitoring. Int J Sports Med. 2024.

3. Greig L, Aspe RR, Hall A, Comfort P, Cooper K, Swinton PA. The predictive validity of individualised load-velocity relationships for predicting 1RM: a systematic review and individual participant data meta-analysis. Sports Med. 2023;53(9):1693-1708.

4. Çetin O, Akyildiz Z, Demirtaş B, Sungur Y, Clemente FM, Cazan F, Ardigò LP. Reliability and validity of the multi-point method and the 2-point method's variations of estimating the one-repetition maximum for deadlift and back squat exercises. PeerJ. 2022;10:e13013.

5. Deboutte J, Alcazar J, Riesbeck M, Walker S, Delecluse C, Van Roie E. Validity of the individualized load-velocity profile to predict one-repetition maximum on a pneumatic leg press device in adults aged 55-81 years. Exp Gerontol. 2026;220:113174.

6. Puente-Alcaraz C, Herrera-Bermudo JC, Yáñez-García JM, Rojas-Jaramillo A, González-Badillo JJ, Rodríguez-Rosell D. Relationship between velocity loss and repetitions completed during the bench press exercise: comparative effect of generalized vs. individualized load-velocity relationship to determine relative load. J Strength Cond Res. 2026.

7. Hickmott LM, Chilibeck PD, Shaw KA, Butcher SJ. The effect of load and volume autoregulation on muscular strength and hypertrophy: a systematic review and meta-analysis. Sports Med Open. 2022;8(1):9.

8. Zhang X, Feng S, Peng R, Li H. The role of velocity-based training (VBT) in enhancing athletic performance in trained individuals: a meta-analysis of controlled trials. Int J Environ Res Public Health. 2022;19(15):9252.

9. Janicijevic D, et al. Hexagonal barbell deadlift one-repetition maximum estimation using velocity recordings. Int J Sports Med. 2024.

10. Weakley J, et al. Load-velocity profiles for weightlifting exercises. PLoS One. 2026.

11. Balsalobre-Fernández C, Xu J, Jarvis P, Thompson S, Tannion K, Bishop C. Validity of a smartphone app using artificial intelligence for the real-time measurement of barbell velocity in the bench press exercise. J Strength Cond Res. 2023;37(12):e640-e645.