What a calories-burned estimate actually is
Of all the numbers a fitness calculator prints, this is the one whose apparent precision is furthest from what it can support. It is worth taking apart, because once you can see what goes into it, what comes out becomes much easier to read correctly.
The formula, in full
The free Calories Burned tool computes MET × 3.5 × kg / 200 × minutes. Its worked default is a MET of 3.5, a weight of 75 kilograms and 45 minutes, giving 3.5 × 3.5 = 12.25, then 12.25 × 75 = 918.75, then 918.75 / 200 = 4.59375 per minute, then 4.59375 × 45 = 206.71875.
That is the entire calculation. Three inputs, four multiplications, no other information involved.
What a MET actually is
A MET is a metabolic equivalent of task: how much energy an activity uses compared with sitting still. Resting is defined as 1 MET. An activity at 3.5 METs is taken to use three and a half times the resting rate.
One MET is conventionally pinned to an oxygen consumption of 3.5 milliliters per kilogram of body weight per minute, and that is where the 3.5 in the formula comes from. The 200 is not arbitrary either. Multiply METs by 3.5 and by kilograms and you have milliliters of oxygen per minute; divide by 1000 to get liters; multiply by the conventional figure of about 5 kilocalories released per liter of oxygen consumed. Dividing by 1000 and multiplying by 5 is dividing by 200. So the default works out as 918.75 milliliters of oxygen per minute, which is 0.91875 liters, which is 0.91875 × 5 = 4.59375 kilocalories per minute.
Now look at where the MET value came from. It is a single figure assigned to an activity — the tool's own field is labelled with walking at about 3.5 — drawn from averages across many people doing something described in a few words. Two people can both be "walking" at speeds, gradients and efficiencies that have little in common. The MET table has one entry for all of them.
Everything the formula does not know
Because the calculation is linear in weight, two people of the same weight doing the same nominal activity for the same time receive exactly the same number. The formula has no way to distinguish them. It has never heard of fitness level, movement efficiency, terrain, temperature, or how hard the person was actually working. It cannot, because none of that was an input.
The sensitivity to the one judgement call you do make is easy to quantify. Because the output is directly proportional to the MET value, choosing 4.5 instead of 3.5 multiplies it by 4.5 / 3.5 = 1.285714: the default becomes 265.78125 instead of 206.71875, a jump of 59.0625, about 28.6 percent. One step of judgement in an input nobody can pin down precisely, and the answer moves by nearly thirty percent.
There is a second, quieter ambiguity. Because 1 MET is resting, the figure includes the energy the body would have used anyway. Subtracting that baseline — using 3.5 − 1 = 2.5 METs instead of 3.5 — gives 2.5 × 3.5 × 75 / 200 × 45 = 147.65625, which is 1 / 3.5 lower, about 28.6 percent again. Both figures are defensible; they answer slightly different questions, and a single printed number does not say which one it is answering.
Why devices disagree
A wearable is doing the same kind of thing with more inputs: motion data, heart rate, whatever the manufacturer's model chooses to use. But it is still a model producing an estimate from proxies, with its own assumptions, and different makers make different assumptions. There is no reason to expect two of them to agree on the same person doing the same thing, and none of them is reading energy expenditure directly. Measuring that properly means capturing and analysing expired air in a laboratory, which is not what any of these things do.
How to read the output
Read it as an order-of-magnitude figure: enough to tell a fifteen-minute stroll from a ninety-minute effort, not enough to distinguish 207 from 230. The six digits after the point in 206.71875 are the arithmetic being tidy, not the estimate being accurate.
Where the number is genuinely useful is in comparisons made on identical assumptions. Hold the MET value and the weight fixed and vary only the minutes, and the ratio between two outputs is exactly the ratio between two durations — that comparison is arithmetic and it is sound. It is the absolute figure, the one that looks like a measurement, that is carrying all the uncertainty.
The tool says as much itself: it describes the output as an estimate from a standard formula and notes that real expenditure depends on fitness, intensity and efficiency. That is the accurate description. The number is a rough scale derived from a population average, and treating it as a reading off an instrument gives it a status nothing in the calculation earned.