VO2max on Your Wrist: How Accurate Are Wearable Estimates?

Your watch has never measured your VO₂max. Not once. A real measurement needs a laboratory: a sealed face mask, a metabolic cart analysing every breath you exhale, and a treadmill or bike protocol that pushes you to genuine exhaustion. What your wrist shows you is an estimate — a statistical model’s best guess, built mostly from how your heart rate relates to your pace or power during the ordinary workouts you record.

That does not make the number useless — far from it. Used well, it is one of the most valuable fields in your health data, because it tracks cardiorespiratory fitness, which is among the strongest predictors of long-term health researchers have ever measured. But using it well means knowing what an estimate can and cannot do. This article explains how the wrist number is actually produced, what the validation research says about its accuracy, why it sometimes jumps around for no apparent reason, and how to feed it the clean data that makes its trend genuinely worth trusting.

How a Watch Guesses Your VO₂max

The physiology behind the estimate is elegant. At steady, submaximal intensities, heart rate rises roughly in proportion to workload — and the fitter you are, the more work you produce at any given heart rate. Picture two runners at the same easy pace: one cruises at 140 bpm, the other needs 170 bpm to hold it. The first runner is working at a smaller fraction of their maximum, which implies a higher ceiling. Your watch observes exactly this relationship during recorded workouts — heart rate from the wrist sensor, workload from GPS pace or cycling power — and combines it with your age, sex, and body weight. From that heart-rate-to-workload curve it extrapolates upward to an estimate of your maximum, without ever asking you to go there.

This approach has a respectable scientific pedigree. Long before smartwatches, researchers built “non-exercise” prediction models that estimated VO₂peak from simple inputs — age, sex, self-reported activity, resting heart rate, waist circumference. The Norwegian HUNT study fitted such a model on thousands of adults who also completed real laboratory tests, and showed that fitness could be predicted surprisingly well at the population level from almost nothing. Consumer wearables are descendants of that idea: the same statistical machinery, calibrated against lab-tested reference data, upgraded with real workout recordings instead of questionnaire answers.

Nes, B.M. et al. (2011). “Estimating VO₂peak from a nonexercise prediction model: the HUNT Study, Norway.” Medicine & Science in Sports & Exercise, 43(11), 2024–2030.

What the Validation Studies Actually Show

The most complete accuracy picture comes from a systematic review with meta-analysis by the INTERLIVE network, which pooled the studies comparing consumer wearables against laboratory testing. The headline finding cuts both ways. At the group level, wearable estimates correlate reasonably well with lab-measured values — averaged across many people, the devices get the picture broadly right. At the individual level, however, the error can be substantial: your personal estimate can sit several ml/kg/min away from your true value, in either direction.

Molina-Garcia, P. et al. (2022). “Validity of Estimating the Maximal Oxygen Consumption by Consumer Wearables: A Systematic Review with Meta-Analysis and Expert Statement of the INTERLIVE Network.” Sports Medicine, 52(7), 1577–1597.

The error is not evenly distributed either. Devices tend to be least accurate for the least fit — and the direction of the miss is usually flattering, overestimating low fitness. That is worth pausing on: the people whose health would benefit most from an honest number are the most likely to be shown a rosier one. A reassuring estimate is encouraging, but it is not a clean bill of cardiovascular health.

Devices also disagree with each other. When researchers put several wrist trackers on the same participants and compared them all against laboratory measurement, the devices produced markedly different estimates from identical sessions — some closer to the lab value, some further away. Your number is not portable: comparing your estimate with a friend’s from a different device, or with your own from last year’s model, is comparing two algorithms, not two bodies.

Passler, S., Bohrer, J., Blöchinger, L. & Senner, V. (2019). “Validity of Wrist-Worn Activity Trackers for Estimating VO₂max and Energy Expenditure.” International Journal of Environmental Research and Public Health, 16(17), 3037.

Why Your Estimate Jumps Around

An estimate built from the heart-rate-to-pace relationship is only as good as the heart rate and the pace it sees. Most of the mysterious week-to-week swings in a VO₂max estimate are not fitness changes at all — they are input problems.

Optical wrist sensors are the first weak link: cold skin, a loose band, or vigorous arm movement can make the heart-rate trace erratic, and a stretch of nonsense heart rate quietly corrupts the curve the model learns from. Pace is the second: on a treadmill without a footpod the watch is guessing your speed, and in dense city streets GPS can wander. Terrain skews the relationship too — climbing a hill at a high heart rate and a slow pace looks, to a pace-based algorithm, exactly like being unfit. And your physiology itself adds noise: heat and dehydration push heart rate upward at any given pace, so a hot, underhydrated run reads as a fitness loss even when nothing changed. Stop-start efforts and interval sessions, meanwhile, simply give the model few steady segments to learn from.

So when your estimate drops two points overnight, the right first question is not “have I lost fitness?” but “what was different about that workout?” Fitness changes over weeks; inputs change every day.

The Trend Is the Real Signal

Here is the reframe that makes the estimate useful: much of the individual error is systematic rather than random. Your device, on your body, with your habits, tends to miss in a consistent direction. That makes the absolute number unreliable but the direction of change informative — if your estimate climbs steadily over three months of consistent training, your fitness almost certainly improved, whatever the true starting value was. The bias travels with you; the trend cuts through it.

Reading the trend well is mostly a matter of timescale. Fitness adapts over weeks and months, so give the estimate the same window you would give the training itself: compare this month against last month, not today against yesterday. A sustained shift of a couple of points that holds across many workouts is telling you something real; the same move inside a single week is telling you about the weather, your watch strap, or your sleep.

If you want context for the absolute value, lab-based reference data exists: the FRIEND registry compiled thousands of true cardiopulmonary exercise tests into reference standards by age and sex. Treat those norms as loose context, not a verdict — your wrist number carries its own error, so placing it precisely on a lab-derived percentile chart implies an accuracy the estimate does not have.

Kaminsky, L.A., Arena, R. & Myers, J. (2015). “Reference Standards for Cardiorespiratory Fitness Measured With Cardiopulmonary Exercise Testing: Data From the Fitness Registry and the Importance of Exercise National Database (FRIEND).” Mayo Clinic Proceedings, 90(11), 1515–1523.

This is also all you need to know about “fitness age”: it is your VO₂max estimate mapped onto those same age norms and handed back to you as a year count. If your estimate matches the average of someone a decade younger, your fitness age is a decade below your calendar age. It is a genuinely motivating way to frame the number — watching a fitness age fall feels better than watching millilitres rise — but it is the same estimate in different clothes, and it inherits every caveat above.

How to Earn a More Reliable Number

You cannot make a wrist estimate lab-accurate, but you can remove most of its noise. The goal is simple: give the algorithm regular, clean examples of the heart-rate-to-pace relationship it is trying to learn.

• Record steady outdoor sessions regularly. Sustained, even-paced runs or brisk walks in the moderate-intensity range, outdoors with good GPS, are exactly the calibration data the model needs. One or two a week keeps the estimate current.
• Wear a chest strap if you can. A chest strap reads the heart’s electrical signal rather than blood flow at the wrist, and feeds the model a far cleaner trace — especially in cold weather or fast arm-swing sports.
• Keep conditions consistent. Flattish routes, similar times of day, reasonable temperatures, and normal hydration. Save the hill repeats and heat sessions for training benefit; just know they are poor calibration runs.
• Keep your profile honest. VO₂max is expressed per kilogram of body weight, so a stale weight entry shifts the estimate. Update your profile when your weight changes meaningfully.
• Ignore single-workout jumps. A two-point move after one run is noise. Judge the estimate on its four-to-twelve-week trend, the timescale on which fitness actually changes.
• Never compare across devices. Different devices produce systematically different estimates for the same body. Pick one, keep it, and read its trend.

Where Lamplit Fits In

Everything above points to one practice: watch the months-long trend, in context, and ignore the daily wobble. That is precisely what Lamplit is built for. With Lamplit Pro, the VO₂max estimate your watch already produces syncs automatically from Apple Health on iOS or Health Connect on Android into your biomarker trends — so instead of a number buried in a watch app, you see its trajectory over weeks and months, right next to the training, sleep, and recovery data that explain it. Lamplit does not compute its own estimate; it makes the one you have readable.

Context is where the estimate becomes actionable. Logging your workouts, sleep, and how you feel is free, and that record is what turns a wobbly line into a story: the interval block that preceded a genuine climb, the stressful low-sleep fortnight behind a dip, the steady base weeks quietly paying off. Genie, Lamplit’s AI coach grounded in peer-reviewed research, can help you read that story instead of reacting to single data points — which, with a noisy estimate of a vital number, is exactly the discipline that pays.

Important Caveats

A wearable VO₂max estimate screens and hints; it does not diagnose. It is least reliable exactly where medical stakes are highest — in people with low fitness, heart conditions, or medications that alter heart rate, where the model’s assumptions bend furthest. If you need a number you can act on clinically — before surgery, in cardiac rehabilitation, or because symptoms worry you — that is a laboratory test ordered by a clinician, not a watch reading. And if you have been sedentary or have a heart condition, talk to a clinician before chasing your estimate upward with hard training. The wrist number is a compass, not a medical instrument.

The Bottom Line

Your watch estimates VO₂max from the relationship between your heart rate and your pace or power, calibrated against lab-tested models. The research verdict is consistent: good at the group level, imperfect for individuals — sometimes by several ml/kg/min, most often flattering the least fit — and different from device to device. None of that ruins the number, because its real value was never the absolute figure. Feed it steady, well-recorded outdoor sessions, keep conditions consistent, and read the three-month trend: an imperfect estimate of one of the most vital numbers in your health is still well worth tracking — and, unlike most health metrics, this one moves when you train.

Start free on Lamplit — log your training and recovery, and with Pro, watch your VO₂max trend sync from your watch into one clear picture of your fitness.