A number appears on the Health app screen every morning. Resting energy. It sits between heart rate and step count, rendered with the same decimal confidence as everything else the Apple Watch tracks — clean digits, no margin of error, no hint that anything about it might be dramatically wrong.
That confidence was tested. A meta-analysis pooling 56 Apple Watch studies compared what the watch reported for resting energy against laboratory-grade calorie measurements — equipment that captures the precise exchange of oxygen and carbon dioxide with every breath.
How Accurate Is Apple Watch Resting Energy?
Resting energy is the Apple Watch’s least accurate calorie measurement — off by an average of 43.3% in a meta-analysis of 56 studies. Everyday activity produced 18.2% error. The heart rate sensor reads accurately, but the algorithm converting heart rate to resting calories lacks the motion data it needs to match any individual’s actual metabolism.
— Choe & Kang 2025 · Physiological Measurement · 56 studies pooled
The resting condition — where the body is still, the heart rate is steady, the measurement should be simplest — carried the largest calorie error of any activity level tested: 43.3%.
During everyday activity, the same watch came closer: 18.2% error. During exercise, closer still. The simplest physical state to measure produced the worst calorie estimate.
On a resting estimate of 1,800 calories, a 43.3% average error means the watch is typically off by around 780 calories in either direction — enough to place someone in a deep surplus or a steep deficit depending on which way the number lands.
Inside the watch, the heart rate sensor is accurate to within 4.43% of medical-grade equipment. Well inside acceptable limits. The sensor reads the pulse correctly. What fails is everything after.
An algorithm converts that heart rate into calories. During a walk or a run, the algorithm draws on a second data stream — the motion sensor — which supplies stride length, arm swing, pace changes. Those motion signals help the math land closer. At rest, the motion sensor goes silent. No steps. No arm movement. No data to refine the estimate. The algorithm falls back on population-average metabolic models built from generic inputs: age, weight, sex, height. A statistical profile applied to one specific body.
The sensor works. The math behind the resting calculation does not have enough information to be accurate for any individual.
Even the motion data has a threshold. After 40, the step count error more than doubles and the watch overwhelmingly undercounts — so the signal that is supposed to save the calorie estimate arrives already short.
The group average obscures the individual problem. Averaged across everyone in the studies, the resting energy error was close to zero. That sounds reassuring — until the averaging is unpacked. The watch overestimated for some people by large margins and underestimated for others by equally large amounts. The positive and negative errors cancelled each other in the group statistic. They did not cancel on any single wrist. A near-zero average error coexists with a 43% absolute error because the watch is not consistently wrong in one direction — it is unpredictably wrong in both.
Across studies, the exact magnitude varied. Different watch models, different lab setups, different populations produced different numbers. The 43.3% is pooled from all 56 studies — not a fixed constant for every wrist. Some studies found worse. Some found less bad. None found rest to be the accurate condition.
Every total daily calorie number the watch displays folds this resting estimate into its foundation. The number that felt most stable — the one someone reviewing their calorie balance trusted most — carries the widest error margin of anything the watch reports. That gap reaches into the total daily figure, and into the larger question of how fitness trackers handle calorie measurement — further than rest alone.