The Riven Journal
The science

Velocity-Based Training Without a $300 Device: How Your Wrist Can Do It

Velocity-based training without a barbell sensor: reps slow as you fatigue and a wrist device reads that. Thresholds, Apple Watch validation, limits.

Baraa Bilal
·
17 min read
Velocity-Based Training Without a $300 Device: How Your Wrist Can Do ItRiven · The science

You can do velocity-based training without a barbell encoder. A wrist device, including an Apple Watch, can read how much your reps slow down inside a set, and rep-speed decay is the entire point of VBT. The catch: the wrist is a noisier vantage point than the bar, so treat it as a proximity-to-failure gauge, not a precise meters-per-second readout. That is the honest version, and it is still useful. Most lifters guess. A velocity based training app on your wrist stops you from guessing.

I have coached a lot of people who own zero VBT hardware and never will. They do not need to know their squat moved at 0.41 m/s. They need to know the bar is slowing down and they are closer to failure than they think (how to measure reps in reserve without guessing). That is a far cheaper problem to solve than the marketing around linear position transducers suggests.

What is velocity-based training (VBT)?

Velocity-based training is using rep speed, how fast the bar (or your limb) moves during the concentric, to guide your training instead of just sets, reps and percentages of 1RM. The core idea: at a fixed load, your reps get slower as you fatigue within a set, and that slowdown is measurable (why reps slow down at the end of a set covers the physiology). VBT uses that velocity signal to autoregulate load, manage fatigue and judge how close you are to failure.

In practice it shows up 2 ways. Some lifters use absolute velocity to pick load ("stop the set when this rep drops below 0.30 m/s"). Others use velocity loss, the percentage drop from your fastest rep in the set, to cap volume. Velocity loss is the more interesting one for most people, because it does not require you to know your exact load-velocity profile. It only requires comparing your reps against each other, within the set you are already doing.

That distinction decides whether a wrist is enough, so it is worth being precise about it. Absolute velocity ("this rep moved at 0.42 m/s") needs an accurate speed in real units and a personal load-velocity profile, which is exactly where a wrist sensor's per-rep noise bites hardest and where a barbell transducer's precision is worth paying for. Velocity loss ("my last rep was 25% slower than my first") is a ratio of the device's own readings against each other, so a consistent measurement bias mostly cancels out, and it maps cleanly onto how close you are to failure.

How does velocity loss tell you how close to failure you are?

Velocity loss is the percentage your rep speed drops from the fastest rep in a set, and it correlates strongly with how many reps you have completed out of your true maximum. In Smith-machine work by González-Badillo and Sánchez-Medina, velocity loss tracked the percent of maximum possible reps with an R² of 0.97 on bench press and 0.93 on squat. That is a tight relationship. It means you can estimate "how close to failure am I?" without actually grinding to failure to find out.

Here is why that matters physically. Early in a set, the bar slows mainly because of metabolite accumulation: acidosis reduces how fast your muscle fibers can shorten. That is the steady creep you see in the middle reps. But the fatigue that actually kills force, calcium-handling fatigue inside the fiber, rises non-linearly and explodes in the last couple of reps before failure. So a velocity-loss signal is best positioned exactly where it counts: flagging the effective end-of-set zone. Chris Beardsley has written a lot about this mechanistic split, and it lines up with what you feel: the speed bleeds off gradually, then the wheels come off fast.

One honest caveat I will repeat throughout: velocity loss is load- and exercise-dependent. At the same true reps-in-reserve, a lighter load shows more velocity loss than a heavy one, and you will complete fewer reps to hit a 20% loss on bench than on squat. A fixed velocity-loss percentage does not equal a fixed reps-in-reserve across every lift and load. Use it as a relative, within-set fatigue gauge ("this set's speed dropped X% versus my fast reps, so I am near failure this set"), not as a universal "you have exactly 3 reps left" oracle.

How much velocity loss should you actually train at?

This is where VBT earns its keep, because the research gives a usable map (velocity-loss thresholds explained has the full table). A dose-response meta-analysis of 9 studies and 336 trained men found a reverse-U between velocity loss and 1RM strength gain: gains peaked around 20–30% velocity loss (~13.3 kg at VL20%, ~13.6 kg at VL25%, ~13.2 kg at VL30%) and fell off hard at the extremes, only ~7.8 kg at VL0% and ~9.3 kg at VL50%. Stopping too early leaves strength on the table. Grinding every set to a crawl does not add more; it adds fatigue.

And training efficiency, strength gained per rep performed, drops linearly as velocity loss rises: about 0.11 kg/rep at VL0%, 0.03 kg/rep at VL30%, 0.02 kg/rep at VL50%. Low velocity loss buys near-equal strength for far fewer reps.

The strength-versus-size split is the practical takeaway. The largest autoregulation meta-analysis (308 participants) found velocity-loss thresholds ≤25% produced significantly greater 1RM strength (MD 2.32 kg) while >25% produced significantly greater muscle cross-sectional-area growth (MD 0.61 cm²). Lower loss biases toward strength and power; higher loss biases toward hypertrophy through accumulated volume.

Pareja-Blanco et al. (2017) showed it in an 8-week squat study: VL20% and VL40% built similar strength, but VL20% did it with 40% fewer reps and gave a bigger jump improvement (CMJ +9.5% vs +3.5%). The VL40% group got more thigh hypertrophy, but cut their percentage of fast type-IIX fibers nearly in half. More slowdown is not simply "working harder." It is a trade.

So there is no single correct number. The right velocity loss depends on your goal (how close to failure should you train?). Which is exactly why I think a failure-proximity readout is more honest than prescribing one universal threshold: it tells you where you landed, and you decide whether that served strength or size.

Do you need a barbell sensor for VBT?

No. A bar-mounted linear position transducer is the gold standard for precision, but it is not the price of admission. The reference units cost real money: a GymAware RS linear position transducer runs around $1,500, plus a yearly software fee of a few hundred dollars on top, and even the cheaper GymAware FLEX is about $495. That hardware barrier is the whole reason most lifters never touched VBT.

These devices are genuinely excellent. An LPT works like a stopwatch and a tape measure: a tether reels off a known distance over a known time, and the device computes velocity directly. That is why it is the criterion researchers trust. But look at who it is for. A college strength coach running 15 platforms wants per-rep precision and a shared dashboard, and $1,500 a platform is a rounding error in that budget. A lifter who trains alone 3 or 4 days a week mostly wants to stop overshooting and undershooting sets, and the question they are asking ("how close am I?") does not need the tether.

The cheaper path is an inertial sensor, the same accelerometer-and-gyroscope chip in your phone and watch. A systematic review of IMU validity found IMUs are mixed against the LPT gold standard, but several (the VmaxPro, and a bar- or wrist-mounted Apple Watch) reach acceptable mean-velocity validity around r ≈ 0.95 at a fraction of the cost. "Acceptable for autoregulating your own training" and "lab-grade" are different bars. For deciding whether your set slowed down meaningfully versus your fast reps, an inertial sensor clears the first bar comfortably.

Can an Apple Watch do velocity-based training?

Yes, for the job most lifters actually want VBT for: knowing how close a set got to failure. Worn on the wrist, an Apple Watch tracks mean barbell velocity almost as well as a clip-on sensor when both are checked against lab motion capture (the numbers are in the next section). It is weaker on the single-moment metrics, peak and propulsive velocity, because the wrist rotates under the bar in ways the bar never does. For velocity loss, which compares a set's reps against each other, that is the right trade: mean velocity is the metric, and the wrist reads it well. No clip, no tether, no 4th piece of gym hardware to charge.

And "eyeballing it" really is the weak link. In a study of coaches judging velocity loss by eye, experienced coaches averaged 2.6 reps of absolute error, up to 20 reps off, and got worse at lighter loads and lower thresholds. If trained coaches cannot see it reliably, the lifter mid-set definitely cannot. Almost nobody deciding on a wrist tool is comparing it to a GymAware; they are comparing it to guessing, and guessing loses.

That is the gap a wrist app is built for (how an Apple Watch detects muscle failure has the longer version). Riven is the Apple Watch app that scores muscle failure. It reads your wrist motion and stays quiet during the set; about three seconds after you rack the weight you get the exercise, the rep count and a 0-100 score. Reps slow down as a muscle nears failure, and that is what the score is built on. No barbell clip, no camera, no extra hardware. It shows you which set was the real one.

What does the validation evidence say: wrist vs motion capture?

The evidence says a wrist-worn Apple Watch measures mean barbell velocity with validity statistically equal to a barbell-mounted device, when both are checked against optical motion capture. A 2023 validation study in the journal Sports tested the Apple Watch Series 7 on the free-weight back squat against a Vicon 3D motion-capture system, the lab criterion, across 22 participants and 547 paired wrist-worn reps. For mean velocity, the wrist-worn watch posted r = 0.952–0.965 with a standard error of estimate of 0.064 m/s (about 10.4%). The barbell-mounted commercial sensor it was compared against (an Enode Pro) managed r = 0.959–0.971 with an SEE of 0.059 m/s. The authors called those an "equal level of validity" for mean velocity.

That is the headline that matters. Where the wrist falls behind is peak and propulsive velocity, the single-moment metrics, where barbell mounting wins because it is not being whipped around by a rotating wrist. The watch's peak-velocity correlation dropped to r = 0.922–0.944 with nearly double the error (0.114 m/s); propulsive velocity held up better, at r = 0.964. If your sport demands peak bar speed off the chest to the second decimal, mount the sensor on the bar. If you want to know whether reps are slowing down across a set, mean velocity is the metric, and the wrist tracks it.

The same study is honest about the wrist's other weakness, and so am I. The same watch mounted on the bar was slightly more precise than on the wrist (mean-velocity SEE 0.049 vs 0.064 m/s), and the wrist dropped more reps, 49 missed versus 19, which the authors pinned on the arm moving during the lift. The wrist is a noisier vantage point than the bar, not just a less accurate one. That is a structural reality, not a software bug, and it is why a wrist tool should be read as a trend across the set rather than a per-rep instrument.

Limitations of wrist-based VBT

I will not oversell this, because a skeptical lifter will spot it instantly. 3 real limitations:

A wrist reads a smaller velocity drop than a bar-mounted sensor at the same fatigue. Wrist motion includes joint-angle change and arm-segment rotation that do not map 1:1 to bar displacement. The fix is not to claim parity with a transducer; it is to read the wrist as a proximity-to-failure gauge, not as a velocity meter.

Velocity loss is unreliable session-to-session for free-weight squats specifically. This is a published critique I am not going to hide: a European Journal of Applied Physiology paper called VL a "flawed method" for the free-weight back squat, because the percent of reps completed at a given threshold swung wildly between sessions (74.8% vs 58.7% of reps at 90% 1RM, VL20%). The path of a free barbell wanders. Machine, cable and Smith-machine movements, where the bar path is constrained, behave far more predictably, and that is where wrist-based proximity estimation stands on firmer ground.

A fixed VL% is not a fixed reps-in-reserve. Already said it; saying it again because it is the most common misuse. Load-dependence means 20% loss on a light pump set and 20% loss on a heavy triple are not the same distance from failure. A 2025 study of 2,972 reps across 19 trained lifters found bar velocity explained only about 30% of the variance in how many reps people felt they had left, and the relationship shifted by exercise, by load and by set number within a session. The authors' phrase: velocity and perceived effort are "complementary, but not interchangeable." Use the signal alongside feel, not instead of it.

Given all that, the defensible claim for a wrist tool is not "lab-accurate VBT." It is an objective, on-wrist read on effort that beats guessing, which is the bar most lifters are actually clearing, since they have no measurement at all. Not a replacement for your encoder if you own one, but a real read for the 99% who do not on whether they hit failure instead of a feeling.

Who should still buy a barbell tracker?

Buy the LPT if you need per-rep precision, peak velocity or shared data across many athletes. That is a coach's tool, not a lifter's. Specifically, a barbell device earns its price if you are:

  • a strength coach running multiple platforms who wants one dashboard;
  • a competitive powerlifter or weightlifter peaking off precise bar speed to the hundredth;
  • in a sport-science setting where absolute velocity drives load prescription;
  • someone whose lifts are non-vertical or rotational enough that wrist motion garbles the signal.

For everyone else, the lifter who mostly wants to stop leaving sets 2 reps short or grinding 3 reps past the point of useful fatigue, the wrist is the right tool, because the question you are asking ("how close am I?") is the question the wrist answers well. And remember the accuracy gap you are actually closing: trained lifters who stop at perceived failure tend to underestimate by 1–2 reps, and beginners are off by more. A noisy objective signal that is directionally right still beats a confident gut feeling that is several reps wrong. If you are not sure which of those describes you, that uncertainty is the whole argument for measuring it.

FAQ

Is velocity-based training worth it for a regular lifter?

If you tend to stop sets short or pile on junk volume, yes. The value is not exact m/s numbers; it is an objective check on how close each set got to failure, so you train in the productive 20–30% velocity-loss range instead of guessing.

Do you need a barbell sensor for VBT?

No. A bar-mounted LPT is the most precise option but costs $300–1,500+. Validated inertial sensors and the Apple Watch reach r ≈ 0.95 for mean velocity at a fraction of the cost, enough to track your own velocity trend within a set.

Can an Apple Watch measure barbell velocity?

Reasonably well. Worn on the wrist, an Apple Watch Series 7 tracked back-squat mean velocity at r = 0.952–0.965 (standard error 0.064 m/s) against 3D motion capture, statistically equal to a bar-mounted sensor. It is weaker for peak velocity (r = 0.922–0.944, error 0.114 m/s) and dropped more reps than a bar mount because the arm moves. It is a proximity-to-failure trend tool, not a precise meter.

Should I track absolute velocity or velocity loss on my wrist?

Velocity loss. It is a ratio of the device's own readings against each other, so a consistent measurement bias largely cancels out, and it maps onto how close you are to failure. Absolute velocity in real units demands precision and a personal load-velocity profile, which is where a wrist sensor is weakest.

What velocity loss should I train at?

Roughly ≤25% biases toward strength and power with less fatigue; >25% biases toward hypertrophy via more volume. Strength gains peak around 20–30% loss and fall off at both extremes. Pick by goal, not by a single universal number.

Who should still buy a barbell tracker?

Coaches running several platforms, competitive lifters peaking off precise peak bar speed, and sport-science settings where absolute velocity drives load prescription. For the lifter who wants to stop leaving sets short, the wrist answers the question they are actually asking.

How is a failure score different from a rep-counting app?

Rep counters tell you how many reps you did. A failure score tells you how hard those reps were, which is the number that decides whether the set built anything. Riven is the Apple Watch app that scores muscle failure; it scores each set the moment you rack it.

Sources

Baraa Bilal
Founder of Riven. Writes about measurement, training, and the small honest signals that separate effort from results.
Read next
The scienceWhat Velocity Loss % Should You Stop a Set At? Strength vs Hypertrophy Cutoffs16 min read The scienceWhy Your Reps Slow Down at the End of a Set (and What It Actually Means)15 min read ProductCan Your Apple Watch Detect Muscle Failure? Yes — Here's How It Works12 min read