Velocity Based Training: What Works, What Does Not, and How to Start

Velocity based training means letting the speed of the bar decide what you lift and when you stop. A phone or a small sensor measures how fast each rep moves. That number, rather than a percentage on a spreadsheet, runs the session.

The method is sold on one promise. Measure your bar speed, work out what your one-rep max is worth today, and adjust the load to match how you actually feel. Most of the device companies on the first page of search results tell some version of that story.

That promise is the weakest-supported part of velocity based training.

The strongest-supported part is something else. It is using the drop in bar speed inside a set to decide when the set is finished. That use has replicated trials and a meta-analysis behind it. It also costs nothing.

This guide is for the coach or lifter deciding whether to buy in. It separates the two. What the research supports, what it does not, and a four-step way to start this week without buying anything.

What velocity based training actually is

Velocity based training uses bar speed as the measure of effort, instead of a percentage of your one-rep max.

It works because of one stable fact. For a given exercise, how fast you can move a load is tightly tied to what share of your maximum that load represents. Light weights move fast. Heavy weights move slowly. The relationship is far more precise than that sounds.

In a bench press test with 120 strength-trained men, bar speed tracked the percentage of one-rep max almost exactly, with an R-squared of 0.98. The same test in a machine-guided back squat with 80 men produced an R-squared of 0.96. Relative strength did not change the picture. Stronger and weaker lifters hit the same speeds at the same percentages of their own maximum.

There is a second detail in that bench press work that matters more than the headline. The lifters who retested after six weeks of training were 9.3% stronger. The speed matching each percentage of their max did not move. Your load-velocity relationship is yours, and it holds while you get stronger. That is what makes a speed reading interpretable at all.

Two terms show up on every device screen and are worth thirty seconds. Mean velocity is the average speed across the whole lifting phase. Mean propulsive velocity ignores the section near the top where the bar is already slowing down. Devices report one or both. Pick one and stay with it, because the numbers are not interchangeable.

One condition runs underneath all of it. Every rep has to be pushed with maximal intent. If you deliberately grind a light weight, the speed reading stops meaning anything. That is a habit most experienced lifters already have on their heavy sets, and a habit most people have to build on their lighter ones.

The part that works: velocity loss

Velocity loss is the simplest idea in the method. Take the speed of the fastest rep in your set. Keep going until a rep comes in a set percentage slower than that one. Then stop.

A 20% velocity loss cap means the set ends when a rep lands 20% below your best. Not at a rep count. Not at failure. At a speed.

This is where the evidence is strongest, and where the findings are genuinely surprising.

A review pooling 19 training studies found that the velocity loss threshold did not influence strength or muscle endurance gains. It influenced almost everything else. Higher velocity loss was better for hypertrophy. Lower velocity loss was better for jumping, sprinting, and speed against submaximal loads.

The individual trials show the same shape. An eight-week squat study comparing a 20% cap against a 40% cap found similar strength gains in both groups. The 20% group improved its countermovement jump by 9.5% against 3.5% for the 40% group. It did that while performing 40% fewer repetitions. The 40% group built more thigh muscle, and also shifted its fastest fibre type toward a slower one. The 20% group kept it.

Read that again. Fewer reps, the same strength, a better jump.

A squat study running caps of 10%, 30% and 45% at moderate loads found all three improved strength and muscle endurance. The 10% group improved sprint performance the most. A bench press study running caps of 0%, 15%, 25% and 50% found the split from the other direction. The 50% group built the most cross-sectional area. The lower caps produced the neuromuscular adaptations. No group beat the others on strength.

So the choice of cap is not a choice about how strong you get. It is a choice about what the strength costs you and what else you keep.

If you take one line out of this article, take this one. Strength does not seem to depend much on where you cap velocity loss. Almost everything else does.

That is also why velocity loss is worth more to a coach than a load calculator ever was. It ends sets on fatigue rather than on arithmetic. An athlete having a poor day stops sooner and does not dig a hole. An athlete having a good day earns the extra reps. If you are building a jump block, a low cap keeps the quality of every rep high, which is the whole point of that kind of work.

The part that does not work as advertised: estimating your max every day

Here is the pitch again. Warm up, measure the speed of three or four loads, fit a line through them, read off today's estimated one-rep max, and set your working weights from that. Auto-regulation, on a screen, in ninety seconds.

The evidence does not support using it that way.

A review that pooled individual data from 20 studies and 434 lifters found the typical prediction error sat close to 10% of the measured maximum. The estimates also skewed high, overshooting the real number by around 3.7% on average. The number of loads used and the choice of modelling method barely changed the result. The review's own recommendation was direct. Test the one-rep max directly wherever it is practical, and treat the velocity estimate as a monitoring tool.

The deeper problem sits underneath every one of those predictions. They all need to know the bar speed at your true maximum. In a free-weight back squat study where lifters tested their max on three separate days, the actual maximum was extremely stable, moving by a couple of kilos. The speed of that maximum was not. It swung around with a coefficient of variation of 22.5%. The paper concluded that this method cannot accurately adjust session loads, because the anchor it depends on will not sit still.

None of that makes velocity useless for tracking. A weekly estimate is fine for watching a trend across a training block. It is not fine for deciding today's squat weight to the kilo, and it is definitely not worth buying hardware for on its own.

One caution on the testing side. Working up to a true one-rep max is a maximal effort, and it carries the risk any maximal effort carries. Use a spotter or safeties. Use loads you have handled before. Do not test through pain, and see a physical therapist or physician first if something already hurts.

Velocity is a good monitoring signal and a poor load calculator. Use it to end sets, not to pick them.

Does velocity based training beat percentage based training?

Straight answer. On the outcomes most people buy it for, no.

A review pooling head-to-head trials compared velocity based programs against traditional percentage based programs. It found no significant difference in back squat one-rep max, jump, linear sprint, or change of direction. Both approaches worked. Neither pulled ahead.

That is not an argument against the method. It is an argument against expecting the method to do the work for you. A twelve-week percentage based plan run properly will make you strong. What velocity adds is a feedback signal a percentage cannot give you: whether today's set actually went the way the plan assumed it would.

The value is information. If nobody is going to act on the information, the device is decoration.

What you need to start, and what you do not

Much of the first page of search results for this topic is written by companies that sell velocity devices. So it is worth saying plainly. You can start without one.

An AI phone app tested against a linear position transducer on the bench press tracked it closely at both 50% and 75% of one-rep max. Correlations came in at 0.90 and 0.92. The differences between app and transducer were trivial, with a bias under 0.03 metres per second. Repeat-to-repeat variability was a little higher on the app. That matters if you are chasing small changes in a research setting. It does not matter much if you are capping velocity loss at 20%.

Three tiers, in the order most people should walk them.

• A phone. Free or close to it. Enough to learn the habit, build a profile, and run velocity loss caps on one lift.
• A dedicated velocity device. Worth it when several athletes are lifting at once and nobody can film every bar.
• Nothing at all. Reps in reserve and rate of perceived exertion are cruder signals, but they are free and they already work. A crude signal you use beats a precise one you ignore.

The harder constraint is usually not measurement. For a coach running one program across a full roster with a small staff, the thing that breaks first is visibility. Knowing who actually completed the session, and what it looked like when they did, matters more than knowing the bar moved at 0.61 metres per second.

How to start in one week

Four steps. None of them takes more than a single session.

If you do not have a plan to bolt this onto, start with a weekly strength and conditioning template and add the cap to its main lift. Do not rebuild the program around velocity. Add the cap, leave everything else alone, and see what changes.

Two rules make or break the whole thing. Every rep gets pushed with maximal intent, or the numbers mean nothing. And an athlete gets compared only against their own history, never against the person on the next rack.

Common mistakes

• Chasing the number instead of the intent. Slowing a rep down on purpose to keep a set alive defeats the entire method.
• Comparing athletes to each other. A squat at 0.45 metres per second means something different for every lifter. The only useful comparison is to that lifter's own history.
• Comparing across exercises, or across studies. Bar speed at a true maximum sat around 0.16 metres per second in the bench press and 0.32 in a machine-guided back squat. In a free-weight back squat it came in near 0.24, and moved around between sessions. A number that means grinding in one lift means comfortable in another.
• Reading one bad day as a verdict. Velocity moves with sleep, food, warm-up and mood. Four weeks of readings is a signal. One session is noise.
• Buying hardware first. The habit is the hard part, and the habit is free. Build it on a phone, then decide whether the device solves a problem you actually have.
• Rolling it out on every lift at once. One lift, one block, then expand.

The short version

Velocity based training is worth using. It is not worth using for the reason it is advertised.

Skip the daily max estimate. Pick one lift, build your own load-velocity line in a single session, cap velocity loss at 20% to start, and check the trend once a month. That is the whole method. Add it to one lift in the program you already run, and leave the rest of that program alone.