Smart ring accuracy: what every published study actually found

The short version

Smart rings are accurate enough to trust for some things and not for others, and the line between the two is clearer than the marketing makes it look. Here is what the published evidence supports, in plain numbers:

• Resting heart rate at night: good. The best study we found put Oura Gen 4's nightly resting heart rate within 1.94% of a Polar H10 chest strap on average, across 536 nights.
• Overnight HRV: good, with more error. The same study found HRV error of 5.96% for Gen 4 and 7.15% for Gen 3. Error rose at very high HRV values and, in a separate study, in older adults.
• Asleep or awake: solid. A 2026 meta-analysis of finger-worn devices pooled sleep/wake accuracy at 87%.
• Sleep stages: weaker. The same review found light, deep and REM detection noticeably less reliable, and one clinical sleep-lab study put Oura's four-stage accuracy at 53%.
• Heart rate during exercise: unreliable. We found no peer-reviewed study of it. Reviewers comparing rings with a chest strap saw them read high on runs or fall apart on a trail descent.

One caveat sits over all of it: nearly every peer-reviewed study tested an Oura ring. The rings that charge no subscription have a preprint, an in-house pilot, or nothing. That gap is the reason this site exists, and the reason our accuracy hub ends with a way to check your own ring against a chest strap.

What a ring measures, and how

A smart ring has no electrodes touching your chest and no view of your brain. It works from three kinds of sensor on the inside of the band, and everything it reports is either measured by one of them or estimated from them. Oura's Ring 5 product page lays out a typical set (Oura):

• Light sensors. Green and infrared LEDs for heart rate, HRV and respiration; red and infrared LEDs for blood oxygen. This light-based method is called photoplethysmography, or PPG: the ring shines light into the skin and reads how much comes back as blood volume in the finger changes with each heartbeat.
• A temperature sensor. Oura's page says it “measures temperature trends.” Samsung says the Galaxy Ring takes “regular readings on skin temperature changes while you sleep” ( Samsung). Both describe trends and changes, not a thermometer reading.
• An accelerometer. It senses movement, which feeds step counts, activity and the asleep-or-awake call.

Heart rate is the most direct of these measurements: the ring counts pulses. HRV is harder, because it depends on the exact timing between beats, and small timing errors add up. Sleep stages are the furthest from direct measurement. A sleep lab defines stages from brain activity, which a ring can't see, so the ring infers them from movement, heart rate, HRV and temperature together. A 2021 paper on Oura's sleep staging called this a “multi-sensor approach” ( Altini & Kinnunen).

That ladder, from counted to inferred, predicts the evidence below almost exactly: heart rate agrees best, HRV a little less, sleep/wake well, and sleep stages least.

What a ring is checked against

An accuracy number means nothing until you know what the ring was compared with. The studies here use three references:

• ECG (electrocardiogram). Electrodes read the heart's electrical signal directly, so the timing of each beat is exact. It's the lab standard for heart rate and HRV. Cao et al. used a chest ECG at home; Liang et al. used ECG in a lab.
• A chest strap. A strap like the Polar H10 also reads the heart's electrical signal, from the chest. The strongest ring study we found used a Polar H10 as its “criterion measurement,” recording single-lead ECG at 1000 Hz. A chest strap is the reference you can actually buy, and it's what our home test protocol uses.
• Polysomnography (PSG). The overnight sleep-lab test, which records brain activity along with breathing and heart signals. It's the reference for sleep stages. Some studies also use actigraphy, a wrist motion monitor, which is a weaker reference.

A ring checked against another consumer wearable (an Apple Watch, a WHOOP band) has not been checked against a reference at all. It has been checked against a second estimate. Keep that in mind when a brand or reviewer says a ring “matched” something.

Every study, in one table

These are every ring validation study in our evidence library, checked on September 19, 2026. The last two rows are not peer-reviewed and are labeled that way. Each study name links to the paper.

A note on two numbers you'll see a lot. MAPE (mean absolute percentage error) is the average size of the miss, ignoring direction: 2% means the ring was off by 2% on a typical night. CCC (concordance correlation coefficient) measures how closely two sets of readings agree, where 1.0 is perfect. Dial et al. report Gen 4 CCCs of 0.98 for resting heart rate and 0.99 for HRV.

What the heart rate and HRV studies found

The headline study is Dial et al. in Physiological Reports (2025). Researchers tracked 13 adults for 536 nights wearing Oura Gen 3 and Gen 4 rings, a Garmin Fenix 6, a Polar Grit X Pro and a WHOOP 4.0, all against a Polar H10 chest strap. The Oura rings had the highest agreement of the group. For Gen 4, the ring sold as Oura Ring 4, nightly resting heart rate error was 1.94% and HRV error 5.96%; Gen 3's HRV error was 7.15%. It's the closest thing to the question this site asks: how close does a ring get to a chest strap?

Two ECG studies back up the direction of that result. Cao et al. (JMIR, 2022) compared Oura with a chest ECG in 35 healthy adults sleeping at home and concluded the ring “could accurately measure nocturnal HR and RMSSD,” while its more advanced frequency-based HRV measures (LF and LF:HF) had high error. Liang et al. (Sensors, 2024) compared Oura with lab ECG in 92 younger and 22 older adults. Agreement was strong overall, but HRV error was higher in the older group, where more than half of participants exceeded 10% median absolute percentage error. Stricter data-quality filtering and windows of 30 minutes or longer improved accuracy, which is why our home test compares 30-minute-plus windows, not single readings.

Outside Oura, the only chest-strap data we found is the Tao et al. preprint on the RingConn Gen 2: 64 seated athletes, 10-minute recordings, a heart rate bias of -0.42 bpm and an rMSSD bias of 1.62 ms against a Polar H10. Those averages are small. The rMSSD limits of agreement, -10.08 to 13.34 ms, mean a single reading could still land 10 ms or more from the strap. It hasn't been peer-reviewed. Our HRV vs chest strap guide goes through all four studies in detail, and smart ring heart rate accuracy covers resting versus workout heart rate.

What the sleep studies found

Sleep has more studies than heart rate, and they tell a consistent story: rings are good at telling asleep from awake, and much less sure about which stage of sleep you're in.

The largest single study, Svensson et al. (Sleep Medicine, 2024), compared Oura Gen 3 with multi-night ambulatory (portable) PSG in 96 people, scoring 421,045 epochs. Oura detected sleep with 94.4-94.5% sensitivity and about 91.7% accuracy, but its specificity, the ability to spot wake, was lower at 73.0-74.6%. Stage accuracy ranged from 75.5% for light sleep to 90.6% for REM. Robbins et al. (Sensors, 2024) found Oura Gen 3 “was not different from PSG” for wake, light, deep or REM in 35 adults, but that study was funded by Oura and its lead author sits on Oura's Medical Advisory Board. We report it; we weight it less.

Two reviews pull the field together. Khan et al. (OTO Open, 2025) pooled six Oura studies covering 388 people and found no statistically significant differences from PSG or actigraphy on total sleep, sleep efficiency, time awake, time to fall asleep or time in each stage. Jin et al. (J Translational Medicine, 2026) pooled 28 articles on 11 finger-worn devices: 87% accuracy for sleep versus wake, but 0.65 for light, 0.81 for deep and 0.74 for REM. The clinical test was harsher. Herberger et al. (Scientific Reports, 2025) put three rings on patients in a clinical sleep lab: group averages were close (Oura's total sleep time within 12 minutes), but individual nights missed by a lot, and four-stage accuracy was 53.18% for Oura.

Our smart ring sleep tracking accuracy guide covers every sleep study, including the older Altini (2021) and de Zambotti (2019) papers, and what they mean for choosing a ring. If you're shopping on sleep alone, our best smart ring for sleep tracking ranks the options with the fee included.

Brand by brand: what we actually know

Oura: the strongest evidence, for older models

Every peer-reviewed study in the table tested an Oura ring. Gen 4 (Oura Ring 4) has the chest-strap result; Gen 3 has the chest-strap HRV result and most of the sleep-lab work. The ring on sale now, Oura Ring 5, hasn't been through a published study yet. The closest thing is DC Rainmaker's review: he found Ring 5's workout heart rate against a Polar H10 “largely pretty good” for indoor cycling and road running, and saw “no meaningful difference between Oura and other wearables for core sleep stats” (DC Rainmaker). That's one reviewer, not a study. See our Oura Ring 4 review for the validated model.

RingConn: a preprint and two apnea studies

RingConn Gen 2 is the only no-fee ring with a chest-strap comparison of its own, the Tao et al. preprint above. The Jin et al. meta-analysis also includes two RingConn sleep-apnea studies (Wu et al. and Guo et al., both 2024). We cite the review's pooled conclusion, not those papers' separate numbers. RingConn's “90.7% Sleep Apnea Monitoring Accuracy” is its own marketing claim (RingConn). On the other side of the ledger, Gadgets & Wearables found the Gen 2 “consistently reported higher values” than a chest strap on runs. We found no published validation of Gen 3 or Gen 2 Air. Our RingConn Gen 2 review weighs it all.

Ultrahuman: an in-house pilot on a ring you can't buy in the US

Ultrahuman publishes one study: an in-house pilot of the Ring AIR's sleep heart rate in 6 volunteers over 36 sleeps, compared with a SleepImage ring and an Apple Watch. All authors work for Ultrahuman, and it isn't peer-reviewed. The Ring AIR isn't sold in the US, and we found no independent validation of the Ring PRO that is; the5krunner wrote that “Independent testing will be the key test” (the5krunner).

Samsung Galaxy Ring: nothing peer-reviewed

Our searches of the Europe PMC research database on September 19, 2026 returned no peer-reviewed validation study of the Galaxy Ring. That doesn't mean it's inaccurate. It means nobody has published the check, so any claim about its accuracy is Samsung's, not a study's.

Circular Ring 2: nothing found, and a name that confuses

We found no published validation of the Circular Ring 2. The Herberger study tested a ring called “Circul,” which is a different product from a different company; don't let the name carry the result across. Tech Advisor's reviewer also wrote, “I've not tested a smart ring that I really think does a good job of tracking heart rate during exercise” (Tech Advisor).

The limits of the evidence

Good studies still have edges, and the edges matter more to you than the headline. Five to keep in mind:

1. Small samples. The best chest-strap study had 13 people. The Ultrahuman pilot had 6. Hundreds of nights help with the averages, but they don't add new kinds of bodies, skin, fingers or ages.
2. Resting, not moving. Dial et al. measured nights. Tao et al. measured seated athletes after exercise. Nobody in our library validated ring heart rate during a workout, and the reviewers who compared rings with a strap on the move (DC Rainmaker on a trail descent, Gadgets & Wearables on runs) saw it go wrong.
3. Group error is not your error. “No significant difference from PSG” is a statement about averages across many people. Herberger et al. found group averages close while individual nights missed badly. A ring can be right on average and wrong about you.
4. Who paid. Robbins et al. was funded by Oura. The Ultrahuman pilot was written by Ultrahuman staff. We couldn't retrieve funding or conflict statements for several others. None of that makes a result false; it changes how much weight one paper should carry alone.
5. Old models, new algorithms. Most studies tested Oura Gen 3, and Svensson named the exact sleep algorithm version. Brands update firmware and algorithms, and a result on one version doesn't automatically carry to the ring or app you buy today.

The practical upshot: use a ring for trends in your own numbers over weeks, not for any single night's reading, and not for training zones.

Health features are not accuracy claims

Rings now advertise blood oxygen, temperature trends, sleep apnea signals and blood pressure patterns. None of the heart rate or sleep studies above validates those features. What we can say, from the sources:

• No Oura ring feature is FDA-cleared; Oura is pursuing clearance for a blood pressure feature (MedTech Dive). We found no smart ring with an FDA-cleared blood pressure feature. See can a smart ring measure blood pressure.
• The Jin meta-analysis calls rings a trend tool and a triage tool for severe sleep apnea, not a diagnosis. RingConn says Gen 3 is “not a medical device” (RingConn). Samsung has announced that an FDA-cleared sleep apnea risk feature is coming to the Galaxy Ring in fall 2026, per press reports (Android Authority); it wasn't live when we checked. More in smart rings and sleep apnea.
• We found no temperature validation study in our library; our temperature tracking guide covers what each brand claims.

If a number worries you, take it to a clinician. A ring can prompt a conversation; it can't end one.

The evidence, against the fee

Here is the uncomfortable part for anyone who refuses a subscription. The ring with the best evidence is an Oura, and Oura gates most of its insights behind a membership: $5.99 a month or $69.99 a year in the US, which adds up to $209.97 over three years on the annual plan. Without it, Oura says you'll only see the three daily scores (Readiness, Activity and Sleep), ring battery, basic profile, app settings and Explore content (Oura Support). The nightly heart rate and HRV detail the studies measured isn't on that list.

The no-fee rings cost nothing after purchase but come with a thinner file: a preprint for RingConn Gen 2, an in-house pilot for an Ultrahuman ring not sold here, nothing for Galaxy Ring or Circular. Evidence is 25% of our ring score, which is why Oura Ring 4 leads on that single metric and still trails overall once the fee is counted; the weights are on our methodology page. Our flagship, the best smart rings without a subscription, ranks the no-fee options, and Oura vs RingConn puts the fee next to the evidence head to head.

How to check your own ring

You don't have to take a brand's word, or ours. The same comparison the best study made, ring against chest strap overnight, is something you can run at home:

1. Wear a chest strap and your ring together for at least seven nights.
2. Line up the time windows, using 30 minutes or more, as Liang et al. found longer windows improved accuracy.
3. Log the ring's resting heart rate and HRV next to the strap's for the same window each night.
4. Work out the average percentage difference. Against Dial's Gen 4 figures (about 2% for resting heart rate, about 6% for HRV), you'll know whether your ring on your finger is in that range.

The full protocol, with a worked example and what to do if the numbers are off, is in test your ring against a chest strap. For the strap itself, see the best chest strap for HRV. One easy fix first: Oura recommends wearing its ring on the index finger for best accuracy (Oura Support), and our which finger guide explains why fit matters.