How optical heart rate works

Turn a sports watch over and you'll see the LEDs. That's the whole sensor, more or less. A 2020 peer-reviewed paper by Bent and colleagues in npj Digital Medicine describes the method, photoplethysmography, as "an optical method for measuring changes in blood volume under the skin".

I don't test devices and I haven't worn any product named here. Every result below names who measured it, and everything was checked as of 3 November 2026.

A tracker isn't a medical device. Wearable readings are estimates for general fitness and wellness. They are not medical devices unless stated for a specific feature in a specific region, and they don't diagnose or treat any condition. If you have symptoms or concerns, talk to a clinician.

Light in, light back

Each heartbeat changes the blood volume under your skin, and blood absorbs light. The LEDs shine into the skin, a light sensor picks up what comes back, and software turns that rhythm into a figure in bpm.

Green is the usual colour. Bent's team calls green the typical wearable wavelength and notes that optical readings mostly measure how much green light gets absorbed.

The software is the part you can't inspect. Bent's paper says makers publish little about how they evaluate accuracy. Vendors also change algorithms through updates, so a study of one watch at one point in time doesn't describe it forever.

Three ways it goes wrong

Bent's paper names three sources of error: skin type, motion artefacts and signal crossover.

Motion is the one you'll meet most. The sensor shifts on your skin as you move. Signal crossover is the related problem of the sensor reading the rhythm of your movement as your heartbeat. Fellrnr, an independent, non-peer-reviewed site, reports users whose sensors seemed to lock on to their cadence.

That's why a steady jog is easier than a burpee.

What the tests found

The biggest overall picture I can point to is a 2020 peer-reviewed meta-analysis by Zhang and colleagues in the Journal of Sports Sciences: 44 articles, 738 effect sizes and 15 brands, measured against ECG or a chest strap. I read the abstract only, so I haven't checked its funding.

Those averages flatter the devices. High and low readings cancel out, so a small average can hide large single errors, and most of the watches in that literature are discontinued.

Two studies show the spread better:

• Bent 2020 (peer-reviewed, authors declare no competing interests): 53 adults aged 18 to 54, against an ECG patch. Consumer devices (Apple Watch 4, Fitbit Charge 2, Garmin Vivosmart 3, Xiaomi Miband) had a mean absolute error of 7.2 bpm at rest and 10.2 bpm in activity. "Activity" here meant walking up to about 50% of maximum heart rate, which is nowhere near a hard session.
• Lambe 2026 (peer-reviewed, authors declare no competing interests): 38 studies and 1,855 participants, 66% male, Apple Watch models through Series 9 and Ultra 2. Mean bias was -0.27 bpm, with limits of agreement of -7.19 to 6.64 bpm. Most single differences fall inside that range. Series 10 and later aren't covered, and the result says nothing about other brands.

Hard sessions get their own piece: why wrist heart rate fails in intervals and the gym.

Skin tone: not settled

Melanin absorbs green light, which is why skin tone comes up. The evidence is mixed. A 2023 peer-reviewed review by Koerber and colleagues covered 10 studies and 469 participants: 4 found significantly lower accuracy on darker skin, 4 found no effect and 2 were mixed. The authors call it preliminary and inconclusive.

Bent 2020 found no significant difference across skin tones, with 7 to 10 people in each group. A published response by Colvonen in 2021 questioned that sample size and the Fitzpatrick scale used to grade skin. So I won't tell you a sensor works equally on all skin tones, or that it fails on dark skin. Neither claim has support.

Tattoos: a real gap

I found one peer-reviewed tattoo study, and it isn't about a watch. Navalta and colleagues (Sensors, 2025) tested a Polar Verity Sense armband on 25 people against a chest strap. Over tattooed skin at rest, mean absolute percentage error was 22.9%. It was 5.1% while running.

There's a catch. Sensor readings dropped to zero at rest in 9 of the 25 people, and with those points removed the tattooed site met the validity thresholds, at 4.2%.

I found no peer-reviewed tattoo study on a wrist-worn watch. That's the gap, plainly, and anyone quoting a figure for your watch and your tattoo is guessing.

How a chest strap differs

A strap like the Polar H10 reads something else altogether. Polar lists its sensor as "Electrical, chest strap": it picks up the heart's electrical signal through the skin, so there's no light path to disturb.

In a 2017 peer-reviewed study of 50 healthy adults by Gillinov and colleagues (abstract read, funding unchecked), a Polar H7 strap agreed with ECG at rc 0.996. Wrist and forearm devices ran from 0.92 (Apple Watch) to 0.67 (Fitbit Blaze). Those models are about a decade old, so don't read the ranking across to today's watches. The strap's lead over optical sensors is what I'd take from it.

If you want a strap, the heart-rate monitor list goes by session type, and the fit and placement steps are worth trying first. The picker narrows the options by sport.

If you have a heart condition, follow your care team's guidance on training zones rather than a watch reading. Etiwy and colleagues (peer-reviewed, 2019) found optical monitors less accurate than an electrode strap in 80 cardiac rehab patients.

What nobody can tell you yet

Current firmware is the biggest unknown. Most papers name the model and not the software version, and I found almost no independent data on 2025 and 2026 flagship watches. Any confident accuracy ranking of this year's models goes beyond what's been published.