Why your watch overestimates calories burned

Your watch says the session burned 600 calories. It is one of the most confidently displayed numbers in fitness, and one of the least reliable.

This is not a reason to throw the watch away. It is a reason to understand which of its numbers deserve trust — heart rate, largely yes; calories, considerably less so.

What the research finds

Independent validation work has repeatedly found that consumer wearables track heart rate reasonably well while estimating energy expenditure poorly. The most frequently cited example put seven popular devices on 60 volunteers of varied age, size, skin tone and fitness, and compared them against continuous telemetry and indirect calorimetry across sitting, walking, running and cycling. Heart rate held up. Energy expenditure did not — for any of the seven — and the authors ended by recommending caution about using wearable energy-expenditure figures in health programmes at all. ¹

That split is the useful thing to take away. The sensor measuring your pulse is doing a decent job. The number of calories on the screen is not a measurement at all; it is a model sitting on top of that pulse reading, and it is the model that fails.

Other work has gone further and tested devices against doubly labelled water, the reference method for free-living energy expenditure. One study put twelve devices on the same adults simultaneously — five on the waist, five on the wrist, two in a pocket — measured a full day in a metabolic chamber, then followed them through fifteen days of ordinary life. ³ Wearing every device at once removes the usual objection that results depend on which model was tested. The problems persisted.

The direction of error is not consistent between devices or individuals, but overestimation is common, and the magnitude is large enough to matter for anyone using the figure to plan intake.

Where the error comes from

A wearable cannot see the thing it is estimating. Energy expenditure is a metabolic quantity. The device measures pulse and acceleration, then applies a proprietary model to guess at metabolism. Every layer of that inference adds error.

Individual efficiency varies substantially. Two people running the same pace at the same weight can differ meaningfully in energy cost, based on running economy, gait, and training history. A trained runner is more efficient — burning fewer calories for the same distance, which is the opposite of what most people assume.

Heart rate is an imperfect proxy for effort. It rises with heat, dehydration, caffeine, stress and poor sleep, none of which increase energy expenditure proportionally. Anything that raises heart rate without raising work tends to inflate the calorie estimate.

Resistance training is the worst case. Heart rate during lifting is driven partly by intramuscular pressure and breath-holding rather than aerobic demand. Algorithms built for steady-state cardio handle this badly.

The overlap nobody mentions

Here is a subtler issue that applies even to a perfectly accurate estimate, including the MET-based figures our own calculator produces.

If a tool says you burned 400 calories in an hour, that is your total expenditure during that hour. But you would have burned something during that hour regardless — roughly 70 to 100 calories for most adults, simply existing. The genuinely additional cost of the exercise is the total minus what you would have spent anyway.

For a short session this is a meaningful fraction of the total. Our calculator states this in its method note rather than quietly letting the larger number stand.

How to use these numbers anyway

Do not eat back exercise calories at face value. This is the single most common way a carefully planned deficit disappears. If you do eat some back, treat half the displayed figure as a conservative starting point.

Use them comparatively, not absolutely. The estimates are far more useful for answering “is this session harder than that one” than “exactly how much did I burn”. Consistent error still ranks sessions correctly.

Let your bodyweight trend arbitrate. Your scale, averaged weekly, integrates every source of expenditure — including the substantial and largely invisible contribution of incidental daily movement. If the trend is not moving as expected, the answer is to adjust intake, not to recalculate the burn estimate more precisely.

Understand what MET values are. The Compendium of Physical Activities assigns each activity a metabolic equivalent, and it is worth knowing how those numbers were arrived at: roughly two-thirds are measured from published studies, and the remaining third are estimated by analogy with similar activities. ² The Compendium was also built to make self-reported activity comparable between studies — not to predict what one particular person burned on one particular afternoon.

The arithmetic underneath adds a second approximation. One MET is defined as 3.5 millilitres of oxygen per kilogram per minute, a figure taken from a reference adult. Real resting metabolic rate varies with body size, body composition, age and sex, so that constant is systematically off for anyone far from the reference — and tends to run high for heavier people, which is the opposite of what most users would guess.

So: averages built partly from estimates, multiplied by a constant that is itself an average. They take no account of your skill, terrain, or how hard you were actually working. A reasonable starting estimate and nothing more.

The bottom line

Calorie burn estimates — from your watch, from a machine display, or from our own calculator — are approximations built on population averages and indirect inference. Treat them as a rough scale of effort rather than a budget you can spend against.

The scale, averaged over weeks, is the only feedback that reliably tells you whether your energy balance is where you think it is.