What is the most accurate wearable for sleep stages?
Oura ranks first for sleep-stage classification, with Apple second and Fitbit and WHOOP sharing third. The ranking compares how well wearables classify wake, light, deep and REM sleep against polysomnography, rather than combining sleep duration or sleep timing into the same score.
Sleep-stage accuracy by brand
The ranking gives the most weight to same-protocol four-stage comparisons against polysomnography. Studies that report stage duration, use different stage groupings or test brands in separate participant groups support confidence but do not override direct epoch-by-epoch comparisons.
Rank | Brand | Verdict |
|---|---|---|
1 | Best overall across the strongest modern PSG comparisons | |
2 | Consistently near the top, with particularly strong REM detection | |
3= | Broad evidence base, but results vary substantially by population | |
3= | Strong modern head-to-head result; clinical evidence is less consistent | |
5= | Competitive in a large clinical comparison, but evidence depth is limited | |
5= | Pixel Watch was competitive clinically, but replication is limited | |
7 | Consistently below the leading group in multi-brand PSG comparisons | |
NR | Useful PSG evidence, but current comparative coverage is too thin | |
NR | Key comparison uses non-equivalent sleep-stage grouping |
How the ranking was decided
Sleep stages are harder to validate than sleep duration because a wearable has to classify every epoch as wake, light, deep or REM without measuring the brain signals used by polysomnography. The strongest evidence therefore comes from studies that put several devices on the same sleepers and compare their epoch-by-epoch classifications with PSG. A recent systematic review and meta-analysis reaches the same broad conclusion: no consumer wearable is consistently superior across every sleep metric or population.
Oura takes first mainly because it leads the two strongest modern healthy-sleeper comparisons. In the 2024 three-device PSG study, Oura Ring Gen 3 reached 76.3% four-stage agreement and Cohen's kappa of 0.65, ahead of Apple Watch Series 8 at 75.0% and 0.60 and Fitbit Sense 2 at 70.9% and 0.55. The study was Oura-funded, so it cannot carry the ranking alone.
The multi-night head-to-head preprint points in the same direction. Oura Ring Gen 3 was again first at about 72.5% accuracy and 0.59 kappa, Apple Watch Series 9 was essentially level on raw accuracy but slightly lower on kappa, WHOOP 4.0 followed, then Fitbit Charge 6 and Garmin. Because the preprint has industry links, it is treated as supporting evidence rather than a decisive result.
Apple takes second because the independent evidence strengthens that pattern. In the 2025 six-device PSG study, Apple Watch Series 8 had the highest overall Cohen's kappa at 0.53 and the best REM classification. That study also explains why Fitbit and WHOOP share third: Fitbit Sense and Charge 5 scored above WHOOP overall, while the newer multi-night comparison reverses that order. The evidence does not justify pretending one clearly beats the other.
The 75-person clinical comparison is a useful stress test because it produces a different order. Fitbit Sense 2, Galaxy Watch 5 and Pixel Watch all landed around 0.40 to 0.42 kappa, while Oura and Apple were lower in that clinical sample. That variation is important rather than inconvenient: sleep-stage performance depends on population, sleep fragmentation and the specific algorithm version.
Garmin sits below that group because its results are repeatedly weaker in multi-brand PSG studies. Older six-device laboratory evidence also places Garmin below Oura and WHOOP. Polar has useful older and separate-group evidence from studies including the 2026 home-PSG comparison, but not enough current same-subject linkage for a numbered position.
Withings is also left unranked for a methodological reason rather than a lack of data. In the 2025 six-device study, ScanWatch combined N3 and REM into a single deep-sleep category. That is not the same four-stage output used by Apple, Fitbit, WHOOP and Garmin, so putting its kappa directly into the leaderboard would create a false comparison. Newer evidence such as the 2026 young-versus-older study is used to judge robustness, not to force incompatible metrics onto one scale.
Supporting studies
OuraOura Ring Gen 3, Oura Ring family, Oura Ring Gen 29 studies
Oura takes first because it leads the two strongest modern healthy-sleeper head-to-head comparisons. Its advantage is not universal: performance drops in clinical and older-adult samples, and one key comparison was Oura-funded while another is an industry-linked preprint.
- Home PSG February 9, 2026Beyond the Hype? A Standardised Real-World Evaluation of Consumer Sleep Trackers (CST) in Extracting Sleep
Oura Ring Gen 3 achieved about 72.5% four-stage accuracy with Cohen's kappa around 0.59.
- Clinical January 12, 2026Performance evaluation of consumer sleep-tracking wearables and nearables in healthy young and older adults
Deep-sleep bias was +71.5 minutes with 423.3% MAPE in older adults and +81.0 minutes with 84.0% MAPE in younger adults, showing weak stage-level concordance despite useful overall sleep tracking.
- Review November 10, 2025The Oura Ring Versus Medical-Grade Sleep Studies: A Systematic Review and Meta-Analysis
Pooled evidence showed no statistically significant mean difference for light, deep or REM sleep duration, but generation-specific accuracy varies.
- Clinical March 19, 2025Performance of Wearable Finger Ring Trackers for Diagnostic Sleep Measurement in the Clinical Context
Four-class sleep-stage accuracy was about 53.2% in this clinically heterogeneous sleep-clinic sample.
- Healthy / lab October 10, 2024Accuracy of Three Commercial Wearable Devices for Sleep Tracking in Healthy Adults
Four-stage agreement was 76.3% with Cohen's kappa 0.65. Light, deep and REM sensitivities were 78.2%, 79.5% and 76.0%.
- Home PSG January 26, 2024Validity and reliability of the Oura Ring Generation 3 (Gen3) with Oura sleep staging algorithm 2.0 (OSSA 2.0) when compared to multi-night ambulatory polysomnography: A validation study of 96 participants and 421,045 epochs
Sleep-stage accuracy ranged from 75.5% for light sleep to 90.6% for REM sleep. REM duration was underestimated by about 4.1-5.6 minutes.
- Clinical November 1, 2023Accuracy of 11 Wearable, Nearable, and Airable Consumer Sleep Trackers: Prospective Multicenter Validation Study
Four-stage classification achieved 54.3% accuracy, macro F1 0.5186 and Cohen's kappa 0.3492.
- Healthy / lab August 22, 2022A validation of six wearable devices for estimating sleep, heart rate and heart rate variability in healthy adults
Multi-state sleep-stage agreement was about 61% with Cohen's kappa around 0.43 against polysomnography.
- Home PSG March 22, 2022Performance of Four Commercial Wearable Sleep-Tracking Devices Tested Under Unrestricted Conditions at Home in Healthy Young Adults
Deep-sleep sensitivity was about 0.72 and REM sensitivity about 0.39 in unrestricted home sleep.
AppleApple Watch Series 7, Apple Watch Series 9, Apple Watch Series 8, Apple Watch Series 66 studies
Apple ranks second. Series 8 and Series 9 sit just behind Oura in the two strongest modern healthy-sleeper comparisons, and Apple Watch Series 8 had the highest Cohen kappa in an independent 2025 six-device PSG study. Deep-sleep classification remains a weaker area than REM.
- Home PSG June 1, 2026Performance of three consumer sleep-tracking devices compared with Actigraphy and Polysomnography
Apple Watch Series 7 underestimated light sleep by 73.8 minutes and overestimated REM by 30.3 minutes on average versus PSG.
- Home PSG February 9, 2026Beyond the Hype? A Standardised Real-World Evaluation of Consumer Sleep Trackers (CST) in Extracting Sleep
Apple Watch Series 9 achieved about 72.3% four-stage accuracy with Cohen's kappa around 0.56.
- PSG March 22, 2025A Performance Validation of Six Commercial Wrist-Worn Wearable Sleep-Tracking Devices for Sleep Stage Scoring Compared to Polysomnography
Four-state sleep-stage agreement was fair to moderate; Cohen's kappa for this device was approximately 0.53.
- Healthy / lab October 10, 2024Accuracy of Three Commercial Wearable Devices for Sleep Tracking in Healthy Adults
Four-stage agreement was 75.0% with Cohen's kappa 0.60. REM sensitivity was 82.6%, while deep-sleep sensitivity was 50.5%.
- Clinical November 1, 2023Accuracy of 11 Wearable, Nearable, and Airable Consumer Sleep Trackers: Prospective Multicenter Validation Study
Four-stage classification achieved 56.4% accuracy, macro F1 0.4910 and Cohen's kappa 0.2976.
- Healthy / lab August 22, 2022A validation of six wearable devices for estimating sleep, heart rate and heart rate variability in healthy adults
Using the third-party SleepWatch app rather than Apple's later native sleep-stage algorithm, multi-state agreement was about 53% with Cohen's kappa around 0.20.
FitbitFitbit Sense 2, Fitbit Charge 5, Fitbit Charge 6, Fitbit Sense, Fitbit Charge 4 +3 more11 studies
Fitbit shares third. It has one of the broadest validation records and performs strongly in the 2023 clinical comparison, but it trails Oura and Apple in healthy-sleeper comparisons and its performance changes materially across populations and models.
- Clinical September 1, 2026Fitbit Sense 2 in Hospitalized General Medicine Patients: A Pilot Study Assessing Feasibility and Validity of Specific Measures
REM sleep was overestimated by about 29.4 minutes and agreement was modest (ICC 0.41).
- Home PSG June 1, 2026Performance of three consumer sleep-tracking devices compared with Actigraphy and Polysomnography
Fitbit Charge 5 overestimated light sleep by 57.7 minutes on average versus PSG.
- Home PSG February 9, 2026Beyond the Hype? A Standardised Real-World Evaluation of Consumer Sleep Trackers (CST) in Extracting Sleep
Fitbit Charge 6 achieved about 66.2% four-stage accuracy with Cohen's kappa around 0.47.
- Clinical January 12, 2026Performance evaluation of consumer sleep-tracking wearables and nearables in healthy young and older adults
Deep-sleep bias was +29.3 minutes with 113.2% MAPE in older adults versus -18.3 minutes with 19.1% MAPE in younger adults, showing a marked age-related drop in stage accuracy.
- PSG March 22, 2025A Performance Validation of Six Commercial Wrist-Worn Wearable Sleep-Tracking Devices for Sleep Stage Scoring Compared to Polysomnography
Four-state sleep-stage agreement was fair to moderate; Cohen's kappa for this device was approximately 0.42.
- Healthy / lab October 10, 2024Accuracy of Three Commercial Wearable Devices for Sleep Tracking in Healthy Adults
Four-stage agreement was 70.9% with Cohen's kappa 0.55. Deep-sleep sensitivity was 61.7% and REM sensitivity was about 67.6%.
- PSG July 1, 2024Assessment of Fitbit Charge 4 for Sleep Stage and Heart Rate Monitoring Against Polysomnography and During Home Monitoring in Huntington's Disease
Four-stage sleep agreement was moderate, with reported overall kappa around 0.59; stage-specific sensitivity varied substantially.
- Clinical November 1, 2023Accuracy of 11 Wearable, Nearable, and Airable Consumer Sleep Trackers: Prospective Multicenter Validation Study
Four-stage classification achieved 64.6% accuracy, macro F1 0.5814 and Cohen's kappa 0.4185.
- Home PSG March 22, 2022Performance of Four Commercial Wearable Sleep-Tracking Devices Tested Under Unrestricted Conditions at Home in Healthy Young Adults
Stage detection was weaker than binary sleep/wake detection; deep-sleep sensitivity was about 0.51 and REM sensitivity about 0.47.
- PSG October 1, 2021Validation of Fitbit Charge 2 Sleep and Heart Rate Estimates Against Polysomnographic Measures in Shift Workers: Naturalistic Study
Stage classification varied by stage, with higher agreement for deep and REM than for light sleep.
- Review November 28, 2019Accuracy of Wristband Fitbit Models in Assessing Sleep: Systematic Review and Meta-Analysis
Newer sleep-staging Fitbit models showed sleep sensitivity around 0.95-0.96 and wake specificity around 0.58-0.69.
WHOOPWHOOP 4.0, WHOOP 3.04 studies
WHOOP shares third because the direct evidence does not support separating it cleanly from Fitbit. WHOOP 4.0 beats Fitbit Charge 6 in one modern head-to-head preprint, while the independent 2025 six-device study puts both Fitbit models ahead of WHOOP. Older WHOOP 3.0 evidence is supportive but company-funded.
- Home PSG February 9, 2026Beyond the Hype? A Standardised Real-World Evaluation of Consumer Sleep Trackers (CST) in Extracting Sleep
WHOOP 4.0 achieved about 68.7% four-stage accuracy with Cohen's kappa around 0.54.
- PSG March 22, 2025A Performance Validation of Six Commercial Wrist-Worn Wearable Sleep-Tracking Devices for Sleep Stage Scoring Compared to Polysomnography
Four-state sleep-stage agreement was fair to moderate; Cohen's kappa for this device was approximately 0.37.
- Healthy / lab August 22, 2022A validation of six wearable devices for estimating sleep, heart rate and heart rate variability in healthy adults
Multi-state sleep-stage agreement was about 60% with Cohen's kappa around 0.44 against polysomnography.
- Healthy / lab February 11, 2020Effect of Wearables on Sleep in Healthy Individuals: A Randomized Crossover Trial and Validation Study
Sleep-stage agreement was good overall; REM/dream and slow-wave sleep showed ICC values around 0.85 and 0.74 respectively.
SamsungSamsung Galaxy Watch 5, Samsung Galaxy Watch Active 22 studies
Samsung shares fifth. Galaxy Watch 5 produced a Cohen kappa of about 0.42 in a large clinical PSG comparison, essentially level with Fitbit Sense 2 and only slightly above Pixel Watch. The problem is depth: Samsung has much less directly comparable modern staging evidence than the brands above it.
- Clinical November 1, 2023Accuracy of 11 Wearable, Nearable, and Airable Consumer Sleep Trackers: Prospective Multicenter Validation Study
Four-stage classification achieved 64.9% accuracy, macro F1 0.5761 and Cohen's kappa 0.4177.
- Clinical March 27, 2023Performance of the Fitbit Charge 2 and Galaxy Watch 2 Compared With Polysomnography in Assessing Patients With Obstructive Sleep Apnoea
Sleep staging showed limitations in a clinical OSA population; binary sleep/wake performance was stronger than detailed stage classification.
GoogleGoogle Pixel Watch1 study
Google shares fifth. Pixel Watch reached a Cohen kappa of about 0.40 in the same 75-person clinical comparison, close to Galaxy Watch 5 and Fitbit Sense 2. With little independent replication beyond that study, a higher position would overstate the evidence.
- Clinical November 1, 2023Accuracy of 11 Wearable, Nearable, and Airable Consumer Sleep Trackers: Prospective Multicenter Validation Study
Four-stage classification achieved 63.6% accuracy, macro F1 0.5669 and Cohen's kappa 0.4044.
GarminGarmin Vivoactive 5 / Venu 3, Garmin Vivosmart 4, Garmin Forerunner 245 Music3 studies
Garmin ranks seventh. Vivoactive 5 and Venu 3 reached about 0.41 kappa in the modern head-to-head preprint, but Vivosmart 4 was much weaker in the independent 2025 PSG study and older Forerunner evidence also sits below the leading group.
- Home PSG February 9, 2026Beyond the Hype? A Standardised Real-World Evaluation of Consumer Sleep Trackers (CST) in Extracting Sleep
Garmin Vivoactive 5 and Venu 3 results were pooled, with about 63.4% four-stage accuracy and kappa around 0.41.
- PSG March 22, 2025A Performance Validation of Six Commercial Wrist-Worn Wearable Sleep-Tracking Devices for Sleep Stage Scoring Compared to Polysomnography
Four-state sleep-stage agreement was fair to moderate; Cohen's kappa for this device was approximately 0.21.
- Healthy / lab August 22, 2022A validation of six wearable devices for estimating sleep, heart rate and heart rate variability in healthy adults
Multi-state sleep-stage agreement was about 50% with Cohen's kappa around 0.25 against polysomnography.
PolarPolar Vantage M2, Polar Vantage V, Polar Vantage V Titan3 studies
Polar remains unranked. Vantage V and Vantage M2 have useful PSG evidence, but the studies use older hardware or separate participant groups and do not give Polar enough current same-protocol linkage to place it confidently against the ranked brands.
- Home PSG June 1, 2026Performance of three consumer sleep-tracking devices compared with Actigraphy and Polysomnography
Polar Vantage M2 showed a positive REM-sleep bias of 32.3 minutes versus PSG.
- Healthy / lab August 22, 2022A validation of six wearable devices for estimating sleep, heart rate and heart rate variability in healthy adults
Multi-state sleep-stage agreement was about 51% with Cohen's kappa around 0.28 against polysomnography.
- Home PSG March 22, 2022Performance of Four Commercial Wearable Sleep-Tracking Devices Tested Under Unrestricted Conditions at Home in Healthy Young Adults
Deep-sleep sensitivity was about 0.45 and REM sensitivity about 0.36.
WithingsWithings ScanWatch1 study
Withings remains unranked. ScanWatch was included in the 2025 PSG comparison, but it groups N3 and REM together as deep sleep rather than reporting the same four-state wake, light, deep and REM scheme. That prevents a clean stage-for-stage ranking.
- PSG March 22, 2025A Performance Validation of Six Commercial Wrist-Worn Wearable Sleep-Tracking Devices for Sleep Stage Scoring Compared to Polysomnography
Cohen's kappa was about 0.22, but ScanWatch grouped N3 and REM together as deep sleep, so its staging output is not directly equivalent to the four-state devices.