Shindo What the number actually is#
VO2max — written both ways, VO2max and VO2 max — is the maximum rate at which your body can take up and use oxygen during exercise, in millilitres per kilogram per minute. Measuring it requires a mask, a metabolic cart and a protocol run to exhaustion. It is a real physiological quantity with a real measurement procedure.
What your watch shows is not that. It is an estimate produced by a proprietary model that observes the relationship between your heart rate and your pace or power during ordinary running or riding, and reports where that relationship sits relative to reference data. The vendors are generally clear about this in their documentation and the interface is generally not, which is how a modelled correlation ends up presented as a measurement with a decimal point.
That distinction explains everything that follows. An estimate moves when its inputs move, and its inputs include a great deal besides your aerobic capacity.
How a wrist estimate is produced#
Roughly, and with vendor-specific variation: the model needs efforts above a certain intensity, with a heart rate it trusts and a speed or power it trusts. It looks at submaximal heart rate for a given output, compares this against its reference model, adjusts for your age and weight and sex, and reports a figure. Better implementations weight recent qualifying efforts and smooth across several of them.
Three assumptions are doing the load-bearing work: that your maximum heart rate is what the watch thinks it is, that your pace reflects your mechanical output, and that today’s heart rate at a given pace reflects your aerobic capacity rather than anything else about today. Each of those is defeated routinely.
Six things that move it without you changing#
Heat. The largest single factor. Cardiovascular drift in warm conditions raises heart rate for the same pace by five to fifteen beats. The model reads that as a lower capacity. A hot fortnight can knock several points off, and they come back in autumn without any training having caused either move.
Hills and terrain. Pace on a gradient is not comparable to pace on the flat. Implementations that apply grade-adjusted pace handle this better than those that do not, and none of them handle trail, sand or deep mud convincingly.
The heart-rate source. A chest strap and a wrist optical sensor do not produce the same trace, particularly early in a run and during pace changes. Switching between them produces a step change in the estimate that has nothing to do with you.
Which runs qualified. The model only uses efforts that meet its criteria. A fortnight of easy running followed by one hard tempo means the estimate is now built on a different sample. So can a single race.
Altitude, and the days after travel. Reduced oxygen availability raises heart rate at any given pace. Estimates fall on arrival at altitude and take weeks to recover, and returning to sea level produces the reverse artefact.
Everything that raises heart rate generally. A developing infection, poor sleep, alcohol the night before, dehydration, caffeine, a stressful week. All of these move submaximal heart rate by a few beats, and a few beats is what the model is reading.
Notice that all six move the number in the same direction as the inconvenient answer: almost everything that makes a day harder also makes the estimate lower. That gives the number a mild but persistent bias toward reflecting your circumstances rather than your capacity.
Why the point is not readable and the trend is#
Given all of the above, the day-to-day figure has a noise band that is comfortably wider than any training-driven change you could produce in the same period. Real VO2max moves slowly — in an already-trained athlete, a few percent over months, and often not at all. The estimate moves by a similar amount in a week, for reasons above.
So the daily number is not a signal, and comparing it with last Tuesday’s is not analysis. What can be read is the direction over eight to twelve weeks, provided the inputs stayed roughly comparable: same sensor, similar terrain, similar season. A steady climb over a build block is meaningful. A three-point drop in July is weather.
The same logic applies to fitness age, which is usually the same estimate expressed against age norms. It inherits every source of noise above and adds a comparison to a population, so it moves for the same reasons and means slightly less.
What it predicts, and what it does not#
Even a properly measured VO2max is a weaker performance predictor than people assume. It sets a ceiling, and among trained athletes almost nobody is at their ceiling. What separates athletes with similar maximal capacity is threshold — the fraction of that ceiling they can hold for an hour — and running economy, the oxygen cost of a given pace.
This is why the number is stubborn in exactly the athletes who care about it most. A beginner’s estimate climbs quickly because everything climbs quickly. An athlete five years in sees a nearly flat line while getting genuinely faster, because the improvement is happening in threshold and economy, which this number does not describe.
If you want a metric that tracks the training you are actually doing, watch fitness, fatigue and form and your pace at a known heart rate. Both respond within a block.
Do you need a real test?#
Rarely, for a self-coached endurance athlete. A lab test gives you a defensible number and, more usefully, a properly located threshold and a set of zones anchored to it. If you are going to pay for a lab, the threshold data is the part worth paying for — the capacity number is the headline and the least actionable line on the report.
For most purposes, finding your own boundaries in the field gets you the same practical decisions at no cost, with more observations across the year.
How to use the number in practice#
- Read the trend over months, never the change since last week.
- Discount summer and altitude entirely, in both directions. The mechanism behind both is the same one that distorts pace — see pace in the heat and at altitude.
- Do not change training because it dropped. Nothing in the estimate tells you what to do differently; if you are worried, look at resting heart rate, sleep and form, which at least have known causes.
- Do not compare with other people. Different models, different sensors, different reference data. The comparison is not between two athletes, it is between two vendors.
- If it never moves, that is normal, and it is not evidence that the training is not working.
Shindo does not estimate VO2 max itself — it reads whatever estimate your watch has already written into Apple Health, and charts the trend rather than the daily figure, alongside a forecast of where your form is heading. Both sit in the paid tier; readiness and its drivers, training load, and your weekly plan over your most recent 60 days of history do not. Where an input is missing, the metric is shown as unavailable rather than filled with a substitute value. The methodology explains that rule.
Read next