Wrist vs chest strap heart rate: which should you trust, and when?

How do wrist and chest strap heart rate sensors work?

The two kinds of sensor measure different things. A wrist or other optical sensor uses LEDs and a photodiode to pick up changes in the size of blood vessels under the skin.¹⁴ A chest strap measures the electrical activity produced when the heart muscle contracts, and needs direct contact with the skin.¹⁴ Both descriptions here come from a maker, Polar, so they are the maker’s account, not a peer-reviewed one.

Polar also says optical sensors can be affected by movement and by changes in ambient light, so they can be a little less reliable than an electrical sensor in some conditions.¹⁴ That is a claim from a company that sells both types, and the studies below are a better guide to how big any gap is.

Which is “better” therefore depends on what you want to know. A wrist band is comfortable and easy to wear all day. A strap is a separate item you put on for a session. Polar’s own positioning is that watches suit all-day activity and sleep tracking, and that a chest strap suits intense training.¹⁴

One caution about the evidence. No NHS, British Heart Foundation or NICE page read for this guide compares wrist and chest readings. The comparison studies below come from laboratories in several countries, so this page cannot say how any of it applies to British users in particular.

Which is more accurate? What the studies found

The studies read for this guide are laboratory or training studies, mostly with small groups of young, healthy people. Each was read as an abstract only, so this guide gives the figures the abstracts state and nothing more.

Wrist devices against a reference, by activity

A study of 60 volunteers aged 38 plus or minus 11 years compared seven wrist devices with continuous telemetry while they sat, walked, ran and cycled. The error was lowest for cycling and highest for walking, and six of the seven devices had a median error below 5% on the bike.⁴ The authors concluded that most wrist devices measure heart rate adequately in laboratory activities, though they estimate energy use poorly.⁴

A second study used a Polar H10 chest strap as the reference in 60 young healthy adults. The Apple Watch 6 had a coefficient of variation below 5% across five activities: sitting, walking, running, resistance exercises and cycling. The Polar Vantage V and the Fitbit Sense depended more on the activity, with coefficients of variation of 2.44 to 8.80% and 4.14 to 10.76%.⁵ So even in one test, one watch stayed under 5% and the others did not.

A wrist watch and an upper-arm sensor against a strap

In 16 adults across nine activities, from lying down to high-intensity intervals, a Polar Verity Sense worn on the upper arm had a mean absolute error of 1.43 bpm against a chest strap.⁶ A Polar Vantage V2 worn on the wrist had a mean absolute error of 6.41 bpm and a mean absolute percentage error of 6.82%.⁶ Its reliability varied with position, from 3.64% sitting after exercise to 23.03% lying down, and the authors noted it did better at higher intensities than at lower heart rates.⁶

Intervals and swimming

In 22 people doing a treadmill session, an interval session and an outdoor session, an optical watch (the Mio Alpha) had a mean absolute percentage error of 5.73% against a Polar reference. The authors thought wrist movement during the high-intensity intervals probably reduced its accuracy.⁷

Swimming showed a wide disagreement. In 26 trained swimmers doing front crawl, a Polar M600 watch had limits of agreement against the H10 strap of −53.0 to 33.9 bpm, and the authors called it less accurate. An optical sensor worn at the temple under a swim cap was valid.⁸

Reading the studies together: no single result says “strap good, wrist bad”. The gap depends on the device, the arm position and the activity, and the studies measure it in different ways.

Scenario table: what the studies say by activity

Two rows explain most of the practical advice. Within one study, cycling gave the lowest wrist error and walking the highest.⁴ The swimming and interval studies are where the authors reported the clearest problems,⁷⁸ but the studies used different measures, so they cannot be ranked against each other.

What makes wrist readings go wrong?

The studies and the maker point to a few causes, with some limits on what they can say.

• Movement. Polar says movement and changing light can affect optical sensors.¹⁴ In the interval study, the authors put the lower accuracy down to wrist movements.⁷
• Position. The wrist watch in the 16-person study was far less reliable lying down than sitting after exercise.⁶
• Water. The swimming study found the wrist watch less accurate, while a sensor at the temple was valid.⁸
• Body and skin characteristics. In the 60-volunteer study, error was higher for men, for a greater body mass index, for a darker skin tone and when walking. The authors reported these as associations in a laboratory.⁴ That was a study of seven particular devices, so it does not show that every watch behaves the same way.
• Tattoos. In 25 people, an upper-arm optical sensor on tattooed skin was not accurate at rest (mean absolute percentage error 22.9%, against under 5% on skin without a tattoo).⁹ That was not a wrist sensor, so the result does not carry over to all watches. In the same study, skin tone was the only characteristic linked to accuracy, at rest and when walking, and tattoo features were not significant.⁹

Nothing read for this guide measured the effect of cold or of how tightly a band is worn, so this page says nothing about either beyond the maker’s general caution about movement and light.

A related warning applies to heart rate variability. A systematic review of 18 studies found that the correlation between wearables and ECG for heart rate variability declined progressively as exercise level increased; 16 of the 18 studies used ECG-based devices and only two used optical ones.¹² A 2026 meta-analysis of pulse rate variability from optical sensors advises against generalising its results to sleep, exercise, stress or everyday life.¹³ For a fuller look at that measure, see the guide to heart rate variability.

When is a chest strap worth it?

A strap is most worth considering when the number itself matters and the arm is moving a lot. Polar says an ECG strap suits intense efforts and extreme arm movement, such as racket sports, and follows fast rises in heart rate in intervals and sprints better.¹⁴ That is the maker’s claim and not one of the studies above, though the interval and swimming results are consistent with it.

There is also some evidence from beat-to-beat recording. In ten healthy adults across five activities, the Polar H10 kept an RR interval signal quality of 99.4% in the high-intensity activities, against 89.8% for a Holter ECG. The authors suggested a simple strap of this type might be recommended as the reference for RR interval measurement when intense, high-movement activities are being studied.¹⁰

If your goal is training by heart rate zone, see heart rate zones. If you are comparing trackers more broadly, best fitness trackers covers the choice.

When is a wrist sensor enough?

For general tracking, the studies are reassuring on steady activity. The 60-volunteer study found most wrist devices adequate for heart rate in laboratory activities,⁴ and six of seven were under 5% median error on the bike.⁴ Polar itself describes wrist watches as the better choice for all-day activity tracking and sleep monitoring.¹⁴

At rest, the picture is less tidy, because the one rest-related result for a wrist watch against a strap (the Vantage V2 lying down) was variable.⁶ Treat a single resting reading with some caution, and look at your own pattern over days. The guides to how to measure resting heart rate and normal resting heart rate explain how to take and read one.

For sleep, the single study read found a mean absolute error of 1.16 bpm against polysomnography with a Fitbit Charge 4. It involved 10 patients with Huntington’s disease, so it is a small, special group and not a general figure.¹¹ For what sleep scores and stages mean, see sleep score and REM sleep; for how a daily readiness number is built, see readiness score explained.

After a workout, a wrist reading can also feed recovery measures; see heart rate recovery. For fitness estimates built from heart rate, see VO2 max.

How can you test your own device? A side-by-side session

The studies cannot tell you how your watch behaves on your wrist. You can find out for yourself with a simple test. The routine below is this guide’s own suggestion; no source read gives a protocol, and the timings are only a convenient choice.

1. Wear both devices. Wear the strap and the wrist device together for the whole session, with the strap snug and its electrodes moistened as the maker advises.
2. Warm up and let both settle. Warm up gently and let both devices settle before you start noting numbers.
3. Hold three steady steps. Work through three steady steps, for example an easy pace, a moderate pace and a hard pace, holding each for a few minutes.
4. Note both readings. At the end of each step, note both readings. You can do the same every few minutes in a longer session.
5. Add a short burst. Add a short burst of faster effort if you want to see how the wrist device follows a quick rise.
6. Repeat on another day. Repeat on another day and in another activity, such as a walk, a run or a gym session.
7. Work out the average gap. Work out the average gap, as in the example below.

A worked example (illustration only)

These ten readings are invented to show the arithmetic. They are not from a study and not from any device. Suppose a wrist watch w and a strap c (the reference) gave, in bpm: w = 112, 118, 125, 131, 140, 152, 158, 149, 135, 122 and c = 110, 117, 126, 133, 147, 160, 166, 154, 138, 121.

• Mean absolute error (MAE): add the gaps, 2 + 1 + 1 + 2 + 7 + 8 + 8 + 5 + 3 + 1 = 38, then divide by 10 to get 3.8 bpm.
• Mean absolute percentage error (MAPE): divide each gap by the strap reading, multiply by 100, and average. Here that comes to 2.58%.
• Bias: the average of w minus c is −3.0 bpm, so on average the watch reads below the strap.
• A different method gives a different answer: the MAE of 3.8 bpm divided by the average strap reading of 137.2 bpm is 2.77%, not 2.58%. Always say which formula you used.

The terms bias, MAE and MAPE are the ones used in the 16-person study.⁶ In that example, the two largest gaps (8 bpm each) fall at the highest readings, which is where the wrist-movement problem would be expected to show. You can look for the same pattern in your own numbers.

Heart rhythm concerns and implanted devices: ask your GP or care team

A heart rate sensor is not a heart test. The NHS says palpitations are common and not usually a sign of anything serious, and that strenuous exercise is among the things that can bring them on.¹ If a tracker shows a number that worries you, the useful step is to talk to a clinician, not to rely on a second sensor.

The NHS suggests seeing a GP if palpitations keep coming back or happen more often, if they last more than a few minutes, or if you already have a known heart condition.¹ It also says that an electrocardiogram (ECG) carried out by health professionals, using electrodes stuck to the skin, may be used to find the cause.¹ A consumer strap or watch is not that test. An irregular pulse is one of the symptoms the NHS lists for atrial fibrillation, and it advises seeing a GP if you think you may have symptoms.³

If you have a pacemaker, the NHS describes it as a device put under the skin of the chest that sends out an electrical signal.² After a pacemaker is fitted, the NHS says your care team will give advice on exercising safely and anything to avoid.² The Polar page read for this guide does not say anything about implanted devices, and no manufacturer notice on the subject was found. So ask your care team before you put on a chest strap or any other sensor.

The only British sources used for facts in this guide are the three NHS pages. Everything on comparative accuracy comes from international studies.

Frequently asked questions