6 Smartwatch Health Numbers That Are Wrong and Which Fitness Readings You Can Trust

Resting heart rate is the one number most smartwatches get consistently right. The optical sensors on the underside of the band, which use green LED light to detect blood volume changes in your wrist, work well when your arm is still. A 2019 study published in the Journal of the American Medical Informatics Association tested six consumer wearables and found resting heart rate accuracy within 5% of ECG readings across all devices tested. For detecting trends over weeks, whether your resting rate is creeping up during a stressful month or dropping as your fitness improves, this number is genuinely useful. The caveat: darker skin tones can reduce optical sensor accuracy. A 2021 paper in npj Digital Medicine found that pulse oximetry and photoplethysmography sensors perform measurably worse on people with Fitzpatrick skin types V and VI, which covers a large portion of Indian users. Resting heart rate is still the most dependable metric on the list, but it is not equally reliable for everyone.

The moment you move from rest to a run, wrist-based heart rate tracking starts to drift. Wrist movement creates motion artifact, noise the sensor cannot fully separate from your actual pulse. At moderate intensities, the error is usually small. At high intensities, sprint intervals, HIIT, competitive cycling, it can be off by 20 to 40 beats per minute. A study in the International Journal of Cardiology compared wrist-worn optical sensors against chest-strap ECG monitors during high-intensity exercise and found errors exceeding 15% in nearly a third of readings. If you are training by heart rate zones for performance, a chest strap is the only tool that holds up. Your smartwatch's exercise heart rate is a rough guide, not a training instrument.

Consumer smartwatches measure blood oxygen saturation using red and infrared light. The principle is the same as a hospital pulse oximeter, but the execution is not. Medical-grade oximeters clip to your fingertip, where blood vessels are close to the surface and motion is minimal. A smartwatch sits on your wrist, reads through more tissue, and moves constantly. The FDA in the United States has explicitly cautioned that consumer wearable SpO2 readings should not be used for medical diagnosis. For healthy people with no respiratory conditions, a smartwatch SpO2 reading below 94% is worth paying attention to, not as a diagnosis, but as a reason to get a proper fingertip oximeter reading. During Covid-19, many Indian households bought fingertip oximeters for exactly this reason. That device costs under Rs 500 and is far more accurate than any smartwatch for this specific number. Use the watch reading as a flag, never as a final answer.

Sleep tracking is where the gap between what smartwatches display and what they actually measure is widest. True sleep staging, distinguishing REM from light from deep sleep, requires polysomnography: electrodes on the scalp, eyes, and chin measuring brain waves, eye movement, and muscle tone. A smartwatch has none of those sensors. It infers sleep stages from heart rate variability, movement, and breathing rate. A 2020 study in Sleep Medicine Reviews analysed 22 consumer wearables and found that while total sleep duration estimates were reasonably accurate, sleep stage classification was not reliable enough for clinical use in any device tested. What your watch calls "deep sleep" is a model's best guess, not a measurement. The total hours slept and the broad pattern of sleep consistency across nights are worth tracking. The percentage of REM sleep your watch reports is not a number to optimise.

Stress scores on smartwatches are derived almost entirely from heart rate variability, the variation in time between consecutive heartbeats. Lower variability is associated with sympathetic nervous system activation, which correlates with stress, fatigue, and poor recovery. The association is real. The leap from HRV to a single "stress score" number is where accuracy collapses. HRV is affected by caffeine, hydration, posture, recent exercise, alcohol, and dozens of other factors that have nothing to do with psychological stress. A study in Frontiers in Physiology found that while HRV is a legitimate biomarker for autonomic function, consumer-grade HRV measurements from wrist sensors carry significant error compared to medical ECG-derived HRV. Your watch's stress score is a loose indicator that something in your physiology shifted, not a reliable measure of whether you are actually stressed.

Calorie expenditure estimates from smartwatches are built on algorithms that take your heart rate, weight, age, and sometimes gender, then run them through population-average equations. A 2017 Stanford University study, one of the most cited in this area, tested seven popular fitness wearables and found that energy expenditure estimates were off by an average of 27%, with the least accurate device showing errors above 90%. The algorithms were not built for Indian body compositions, activity patterns, or metabolic profiles. They were built on datasets drawn primarily from Western populations. Treat the calorie number as a rough order-of-magnitude indicator: a 400-calorie workout is probably more than a 150-calorie walk. Whether it was exactly 400 or 280 is unknowable from your wrist.

The pattern across all six metrics is the same: smartwatches measure best when the body is still and the signal is clean, and worst when the body is moving and the inference chain is long. Resting heart rate sits at one end of that spectrum. Calorie burn and sleep stages sit at the other. The watch is not lying to you, it is doing the best it can with a green LED and an accelerometer. The mistake is treating every decimal point on that screen as though it came from a lab.