Smartwatches typically land somewhere in the ballpark for calories burned, but the margin of error is wide enough to cause real problems if you trust the numbers too closely. A large umbrella review of systematic reviews found that wearables underestimated energy expenditure by an average of about 3 percent, which sounds reassuring until you see that individual readings could swing anywhere from roughly 21 percent below to 15 percent above the true value. The accuracy depends heavily on what you are doing, which brand you are wearing, and even your body composition.
What “Accurate” Means in This Context
To evaluate whether a smartwatch gets calorie counts right, researchers need something to compare it against. The gold standard for measuring total daily energy expenditure in real-world conditions is doubly labeled water, a technique where you drink water containing traceable isotopes and researchers calculate your carbon dioxide production over days or weeks from urine samples. Validation studies have shown that this method agrees with room-based calorimetry to within about 1 to 2 percent on average, making it about as close to a “true” calorie count as science can get.1PubMed. Seven-day validation of doubly labeled water method using indirect room calorimetry For shorter exercise bouts in a lab, researchers typically use indirect calorimetry, which measures the oxygen you breathe in and the carbon dioxide you breathe out to calculate energy expenditure in real time.
The problem is that your smartwatch does not have access to either technique. It relies on your heart rate from a wrist-based optical sensor, motion data from an accelerometer, and a proprietary algorithm that folds in whatever personal data you entered during setup: age, sex, weight, height. Some newer algorithms also factor in BMI and interactions between those demographic variables to try to improve predictions.2Scientific Reports. Developing and comparing a new BMI inclusive energy expenditure algorithm on wrist-worn wearables But the gap between what a lab-grade metabolic cart can measure and what a wrist sensor can infer is substantial, and that gap is where the errors live.
Accuracy During Walking and Running
Walking and running are the activities most people track most often, and they are also among the best-studied. A 2022 study that tested Apple Watch, Garmin Forerunner, and Huawei Watch GT against a portable metabolic analyzer during outdoor walking and running found errors that varied a lot by brand. During walking, the average error ranged from about 10 percent for Huawei to 32 percent for Garmin, with Apple Watch sitting around 20 percent. During running, the spread tightened somewhat, with errors of roughly 12 percent for Huawei, 22 percent for Garmin, and 24 percent for Apple Watch.3Frontiers in Physiology. Validity of three smartwatches in estimating energy expenditure during outdoor walking and running Both Apple Watch and Garmin significantly overestimated calories during both activities, while Huawei’s estimates were not statistically different from the reference.
An earlier lab study from 2019 testing Microsoft Band, Fitbit Surge, TomTom, and Apple Watch on a treadmill found similar patterns but with larger errors overall. The best performers across rest, light walking, moderate jogging, and vigorous running were Microsoft Band (about 35 percent average error) and Apple Watch (about 42 percent), with errors lowest during light walking and vigorous running and worst during moderate-intensity jogging.4Journal for the Measurement of Physical Behaviour. Accuracy of Commercially Available Smartwatches in Assessing Energy Expenditure During Rest and Exercise No smartwatch in that study passed formal equivalence testing against indirect calorimetry at any intensity.
The pattern that emerges across studies is that steady-state aerobic activities with rhythmic arm movement tend to produce better readings than activities where your wrist stays relatively still or moves unpredictably. Walking and running give the accelerometer a clean, repetitive signal. Cycling, weight lifting, and activities with lots of arm variation tend to be harder for the algorithms to interpret.
Accuracy During Cycling and Other Activities
Cycling is a good illustration of where smartwatches can struggle. Your wrist barely moves on a handlebar, so the accelerometer contributes less useful data and the algorithm leans more heavily on heart rate. A study testing four low-cost smartwatches during stationary cycling found that two of the devices had small average biases (under 1 kilocalorie), while two others overestimated by 8 to 12 kilocalories per bout on average, with the disagreement growing wider as exercise intensity increased.5PLOS ONE. Validity of four low-cost smartwatches in estimating energy expenditure during cycling in Chinese untrained women The increasing-bias-at-higher-intensity pattern showed up across all devices tested, meaning the harder you pedal, the less you should trust the number on your wrist.
Resistance training, yoga, swimming, and sports like tennis present even more challenges. Arm movements during a bench press look nothing like arm movements during a jog, and most algorithms were originally trained primarily on walking and running data. Swimming adds the complication that water interferes with the optical heart rate sensor. If you see wildly different calorie estimates from your watch during a swim compared to a chest-strap-based estimate, the optical sensor losing contact with your skin through water turbulence is a likely culprit.
The Overestimation Problem
Across multiple studies and brands, the most common direction of error is overestimation. The umbrella review in Sports Medicine, which synthesized findings from dozens of systematic reviews, found that wearables generally underestimated energy expenditure on average by about 3 kilocalories per minute, but the spread of individual errors was enormous, ranging from underestimates of roughly 21 kilocalories per minute to overestimates of about 15 kilocalories per minute.6Springer Link / Sports Medicine. Keeping Pace with Wearables: A Living Umbrella Review of Systematic Reviews Evaluating the Accuracy of Consumer Wearable Technologies in Health Measurement That wide range means your watch could be quite close to reality on one workout and dramatically off on the next.
A 2025 study comparing Apple Watch, Garmin, Samsung, and Fitbit found that Garmin and Samsung systematically overestimated calories, while Apple showed a more modest positive bias of about 23 kilocalories per session. Fitbit was interesting: its average bias looked fine on paper, but that was because it produced some wildly implausible readings (over 450 percent of the true value in roughly 13 percent of trials) that, once included, pushed its average error far above every other brand. When those outliers were removed, Fitbit actually underestimated slightly.7PLOS ONE. Body fat, skin tone, and the accuracy of smartwatch caloric expenditure estimates The takeaway: averages can be misleading. A device that is “accurate on average” might still give you personally a number that is off by a large margin on any given day.
Why Body Fat Percentage Matters
One of the more under-discussed factors in smartwatch accuracy is body composition. The same 2025 study found that higher body fat percentage was associated with larger errors across all tested brands, but the effect was not uniform. Errors climbed more steeply with increasing body fat for Fitbit and Garmin than for Apple Watch.7PLOS ONE. Body fat, skin tone, and the accuracy of smartwatch caloric expenditure estimates
There are a few reasons this happens. Extra tissue between the sensor and the blood vessels on your wrist can weaken the optical heart rate signal, introducing noise into the data the algorithm depends on. People with higher body fat also tend to have a different relationship between heart rate and energy expenditure than leaner individuals performing the same activity, and most consumer algorithms were developed using datasets that skew toward younger, leaner, and more active participants. If you do not look like the people in the training data, the model’s predictions drift further from reality.
The same study examined whether skin tone affected accuracy and found that it did not significantly alter calorie estimates, though darker skin pigmentation has been shown to affect the quality of the underlying heart rate data in some devices. A separate study looking specifically at heart rate accuracy found that darker skin pigmentation reduced data quality for all tested devices but only increased measurement error by about 1 beat per minute for one particular device.8PubMed Central. Influence of skin pigmentation on the accuracy and data quality of photoplethysmographic heart rate measurement during exercise The calorie estimate depends on heart rate but also on accelerometry and demographics, so a small heart rate error does not automatically translate into a proportional calorie error.
Resting vs. Active Calories
Your watch reports two kinds of calorie numbers: resting (or basal) calories, which account for the energy your body uses just to stay alive, and active calories, burned through deliberate movement. Most of the accuracy research focuses on active calories because that is where smartwatch estimates diverge most from lab measurements.
At rest, the variation between different brands is actually at its worst. One study found that the disagreement between different manufacturers’ calorie estimates was greatest during seated rest, with variation among devices reaching about 54 percent, and this spread shrank progressively as exercise intensity increased.4Journal for the Measurement of Physical Behaviour. Accuracy of Commercially Available Smartwatches in Assessing Energy Expenditure During Rest and Exercise This sounds paradoxical: you would think sitting still would be the easiest thing to measure. But at rest, the heart rate signal is low, the accelerometer sees no movement, and the algorithm is essentially guessing based on your demographic profile. Since each manufacturer uses a different formula for that guess, the numbers scatter widely.
For daily calorie totals, the resting estimate matters more than most people realize because basal metabolism accounts for the majority of total energy expenditure. If your watch’s resting estimate drifts by even a few percent, it can add or subtract hundreds of calories over a full day, potentially swamping whatever active-calorie error occurred during your workout.
How People Actually Use These Numbers
The practical consequence of these errors depends on what you are doing with the data. If you use your smartwatch to compare relative effort across workouts, say to notice that Tuesday’s run was harder than Thursday’s, the numbers are useful even if they are systematically off. Trends and comparisons tolerate consistent bias reasonably well.
The trouble starts when people treat the calorie number as literal and make food decisions based on it. Research exploring how wearable device users relate their tracked data to eating habits found that users do make nutritional decisions based on calorie expenditure and activity completion data from their devices, but those decisions can cut both ways: some people eat more nutritious food, while others use a high calorie-burn number as permission to eat more indulgently.9Research Square. Exploring wearable device use in relation to eating habits If your watch overestimates your run by 150 calories and you “eat back” those calories with a post-workout snack, the math works against you over time.
The interaction between technology and goals can also produce unexpected effects. A study on a bite-counting wearable device found that when participants were given a restrictive eating goal, they compensated by taking larger bites, ending up consuming about the same amount as people with a more generous goal.10Journal of the Academy of Nutrition and Dietetics. Effects of Bite Count Feedback from a Wearable Device and Goal Setting on Consumption in Young Adults The broader point applies to calorie tracking too: when people focus on hitting a specific number, they find ways around the constraint, sometimes without even realizing it.
Practical Ways to Get Better Data from Your Watch
You cannot make a consumer smartwatch match a metabolic cart, but you can reduce the noise. Fit matters more than most users think. A loose watch slides around during exercise, disrupting the optical sensor’s contact with your skin and degrading the heart rate reading that the calorie algorithm depends on. Wearing the watch snugly about two finger-widths above your wrist bone, where blood vessels are closer to the surface, gives the sensor a cleaner signal.
Keeping your profile information current also helps. If you have lost or gained weight since you set up the device, the algorithm is still using the old number for every calculation. The same goes for height and age, though those change less frequently. Some devices let you enter body fat percentage or VO2 max data; if yours does, updating those values after a fitness test gives the algorithm better inputs to work with.
Selecting the correct activity type before a workout helps the algorithm apply the right model. “Outdoor run” and “indoor cycle” use different assumptions about the relationship between your heart rate, arm movement, and energy cost. Letting the watch auto-detect your activity or tracking everything as generic “exercise” forces the algorithm to guess, and that guess costs accuracy.
Perhaps the most useful mental adjustment is to treat the number as a rough estimate with an error margin of at least 20 percent in either direction. If your watch says you burned 400 calories, the true number could plausibly be anywhere from about 320 to 480. Building that uncertainty into your expectations, especially if you are using calorie data for weight management, keeps you from over-relying on a number that was never designed to be precise at the individual level.
Why Accuracy Has Not Improved as Fast as You Would Expect
Given how rapidly other aspects of smartwatch technology have advanced, it is reasonable to wonder why calorie estimates are still this rough. A systematic review surveying the evolution of wrist-worn activity trackers noted that the sheer variety in study designs, participant populations, device generations, and statistical metrics used to assess accuracy makes it genuinely difficult to compare results across time or to declare that newer devices are categorically better than older ones.11Frontiers in Digital Health. Advancement in wrist worn physical activity trackers: technological developments and measurement accuracy: a systematic review The hardware has improved, with newer optical sensors sampling at higher rates and accelerometers becoming more sensitive, but the fundamental problem is biological, not technological.
Energy expenditure depends on variables that a wrist sensor simply cannot see: muscle mass, metabolic efficiency, hormonal state, ambient temperature, whether you ate recently, and even your emotional state, which can subtly alter metabolic rate. Two people of the same age, sex, weight, and height can burn meaningfully different amounts of energy performing the same activity, and no wrist-based device currently captures enough of those individual differences to close the gap. Algorithms keep getting more sophisticated, but they are still making educated guesses from a narrow set of inputs about a process that involves your entire body. That is why the errors have plateaued rather than steadily vanishing with each new model year.
When Smartwatch Calories Are Good Enough and When They Are Not
For most recreational exercisers who want a general sense of how active they are, smartwatch calorie numbers serve their purpose. The broad strokes are usually right: a hard run will show more calories than a gentle walk, and an active day will clearly separate from a sedentary one. That level of resolution is enough to motivate consistency, compare workouts, and notice trends over weeks.
Where the numbers become unreliable enough to cause harm is in tight calorie-balance scenarios. If you are trying to create a precise daily deficit for weight loss, or if you are an athlete managing fueling for performance, treating your watch’s calorie output as ground truth can lead to undereating, overeating, or frustrating stalls. In those situations, using the watch for relative comparisons while relying on other methods for absolute calorie accounting, such as tracking body weight trends over weeks rather than daily calorie math, gives you better information with less risk of being misled by a sensor that was designed for convenience, not clinical precision.