9 WEARABLES · PSG-VALIDATED ACCURACY

Which sleep tracker is actually accurate?

No wristband or ring measures your sleep. A lab PSG reads your brain waves; a wearable only estimates stages from heart rate, movement and temperature. Here's how 9 devices really stack up against the lab.

0.65

Best 4-stage agreement (Oura, healthy)

0.75

Where human sleep scorers agree

0

Devices that reliably measure wake

9

Wearables compared vs lab PSG

Build your comparison

Compare sleep trackers head-to-head.

Pick 2–4 devices and see them side by side on the six metrics that matter: staging in healthy and clinical sleepers, deep & REM detection, sleep/wake, and total-sleep-time bias. The best value in each row is highlighted.

Choose devices to compare

Tap to add or remove · 2–4 at a time

Oura Ring

Apple Watch

Fitbit Sense 2

WHOOP

Polar

Garmin

Samsung

Fitbit Charge 5

Withings

Oura Ring wins the most metrics here (3 of 6). These are healthy-adult figures; accuracy drops in older adults and sleep-disorder patients, and no device reliably measures wake.

Get the Kygo app

The full ranking

It's a tie: Apple Watch and Fitbit Sense 2 each lead on 3 of 6 metrics. All of them overestimate sleep and can't reliably measure wake.

Apple Watch wins the most metrics here (3 of 6). These are healthy-adult figures; accuracy drops in older adults and sleep-disorder patients, and no device reliably measures wake.

All nine, ranked on the headline metric.

Sleep staging (light / deep / REM / wake) is the hard task and the number people quote. Here's every device ranked by healthy-adult agreement, with how it holds up in clinical patients. Devices tested only in a mixed or clinical sample come last. Scroll sideways on mobile.

Cohen's kappa vs in-lab PSG · 0.4–0.6 = "moderate," the consumer ceiling · human sleep scorers agree at ~0.75 · green ≥ 0.50. Ranked by healthy-adult agreement; devices tested only in a mixed or clinical sample are listed last. Mixed sample = Schyvens 2025, about half with suspected sleep apnea.

Get the free State of Wearable Accuracy report

Read this before you trust a number

The one thing that decides accuracy.

Who's wearing it matters more than which brand it is. The same device scores far higher in screened healthy sleepers than in older adults or real sleep-clinic patients, which is why the comparison above splits staging into healthy vs clinical.

Screened healthy adults

κ 0.65 best case

Ceiling numbers live here

Young, health-screened sleepers. The manufacturer-funded "best" figures come from this population: a best-case sample, not a method trick.

Older adults & patients

κ 0.30–0.42 real users

Every device drops

Sleep-clinic patients and older adults. Apnea-fragmented sleep is harder to stage, so Oura falls 0.65 → 0.35 and Apple 0.60 → 0.30. Fitbit holds up best (0.55 → 0.42).

A 4-stage kappa of 0.4–0.6 is "moderate" and 0.6–0.8 "substantial." The best consumer devices reach 0.55 to 0.65 in healthy adults; two human PSG scorers agree at ~0.75, so no wearable is "accurate" in an absolute sense. Schyvens 2025 tested a mixed sample (about half with suspected sleep apnea), so its figures appear only as a labelled second figure. Robbins 2024 · Lee 2023

In detail

The full breakdown, tap any device.

Every wearable's independent finding vs a lab PSG, the manufacturer's claim, its per-stage and timing numbers, and what it's best and weakest for.

Oura Ring Gen 3

κ 0.65

What the studies found vs lab PSG

Kappa 0.65 in healthy adults (Robbins 2024, in-lab PSG), the highest of any device, ahead of Apple Watch (0.60) and Fitbit (0.55) in the same study. Across 96 healthy adults over multiple nights it caught 64.0% of deep sleep, 72.6% of REM and 73% of wake, with total sleep 2.1 min short (Svensson 2024). Both studies are Oura-funded. In sleep-clinic patients it drops to kappa 0.35 (Lee 2023) and 53% accuracy (Herberger 2025).

4-stage staging (κ)

Healthy

0.65

Clinical

0.35

healthy figure is Oura-funded (Robbins 2024)

Deep & REM detection

Deep

64%

REM

72.6%

clinical: deep 77.8% · REM 71.2%

Sleep / wake & timing

2-state

n/a

TST bias

−2.1 min

TST: Svensson 2024

Manufacturer claim

OSSA 2.0 staging accuracy 75.5–90.6%; sleep/wake sensitivity ~94% (Svensson 2024, Oura-linked). These are raw % / PABAK figures, not kappa, so they are not directly comparable with the kappas above.

• How it tracks Ring PPG + skin temperature; overnight autonomic + movement staging.
• Best for Sleep staging, total sleep time and catching wake (73%); the vitals measured while you sleep
• Weakest for Sleep-disorder patients (0.65 → 0.35)

View Oura Ring on Amazon

Apple Watch S8

κ 0.60

What the studies found vs lab PSG

Kappa 0.60 in healthy adults (Robbins 2024, an Oura-funded in-lab PSG study), second to Oura, with the best REM detection of any device (82.6%) but only 50.5% of deep sleep caught. In a mixed sample where about half had suspected sleep apnea it scored 0.53 (Schyvens 2025), and 0.30 in sleep-clinic patients (Lee 2023), where REM detection falls to 42.8%.

4-stage staging (κ)

Healthy

0.60

Clinical

0.30

healthy figure is Oura-funded (Robbins 2024)

also 0.53 in a mixed sample (Schyvens 2025)

Deep & REM detection

Deep

50.5%

REM

82.6%

clinical: deep 41.3% · REM 42.8%

Sleep / wake & timing

2-state

93%

TST bias

+8 min

2-state: Series 8 · Robbins 2024 · TST: Robbins 2024

Manufacturer claim

Apple white paper (Sept 2023): ~1,400 nights, kappa 0.63, sensitivity 97.8% / specificity 76%. Its FDA-cleared apnea notifications (Sept 2024) show weighted sensitivity 66.3% (severe-OSA 89.1% but moderate-OSA only 43%).

• How it tracks Watch accelerometer + PPG; wrist-based stage classification.
• Best for REM detection (82.6%); second on sleep staging
• Weakest for Catches only about half of wake (52%), so it over-reports sleep; REM falls to 42.8% in patients

View Apple Watch on Amazon

Fitbit Sense 2

κ 0.55

What the studies found vs lab PSG

Kappa 0.55 in healthy adults (Robbins 2024, an Oura-funded in-lab PSG study), third on staging, with deep 61.7%, REM 67.6% and total sleep within 3 min. It holds up best in sleep-clinic patients: kappa 0.42 (Lee 2023), the smallest drop of any device, with deep 67.1% and REM 68.1%. The original Sense scored 0.42 in a mixed sample where about half had suspected sleep apnea (Schyvens 2025).

4-stage staging (κ)

Healthy

0.55

Clinical

0.42

healthy figure is Oura-funded (Robbins 2024)

also 0.42 for the original Sense in a mixed sample (Schyvens 2025)

Deep & REM detection

Deep

61.7%

REM

67.6%

clinical: deep 67.1% · REM 68.1%

Sleep / wake & timing

2-state

91%

TST bias

−3 min

2-state: Sense 2 · Robbins 2024 · TST: Robbins 2024

Manufacturer claim

Google/Fitbit (2026, in press) report an updated algorithm at kappa 0.63 (up from 0.56) and 77% accuracy (up from 71%). Not yet independently validated.

• How it tracks Wrist PPG + motion; Fitbit Sleep Stages algorithm.
• Best for Holding up in sleep-disorder patients (smallest drop); total sleep within 3 min
• Weakest for Third on staging; still misses about a third of wake

View Fitbit on Amazon

WHOOP 3.0

κ 0.44

What the studies found vs lab PSG

Kappa 0.44 in healthy young adults (Miller 2022, in-lab PSG), with deep 62% and REM 66% caught and total sleep 12.2 min short. WHOOP 4.0 scored 0.37 in a mixed sample where about half had suspected sleep apnea (Schyvens 2025), catching 69.6% of deep sleep there. No sleep-clinic data exists for WHOOP.

4-stage staging (κ)

Healthy

0.44

Clinical

not tested

also 0.37 for WHOOP 4.0 in a mixed sample (Schyvens 2025)

Deep & REM detection

Deep

62%

REM

66%

Sleep / wake & timing

2-state

86%

TST bias

−12.2 min

2-state: WHOOP 3.0 · Miller 2022 · TST: Miller 2022

Manufacturer claim

WHOOP cites internal validation; note the Miller 2022 lab (CQU) receives WHOOP funding. The only other published 4-stage kappa is WHOOP 4.0 at 0.37 in a mixed sample (Schyvens 2025).

• How it tracks Screenless arm strap; PPG + motion, worn 24/7.
• Best for Fourth on staging, ahead of Polar and Garmin; balanced deep (62%) and REM (66%) detection
• Weakest for No clinical data; under-reports total sleep (−12.2 min)

View WHOOP on Amazon

Polar Vantage V

κ 0.28

What the studies found vs lab PSG

Kappa 0.28 in healthy young adults (Miller 2022, in-lab PSG), with deep 33% and REM 49% caught. Total sleep is close (−0.8 min), but it catches only 51% of wake. No sleep-clinic data, and no PSG study of newer Polar models.

4-stage staging (κ)

Healthy

0.28

Clinical

not tested

Deep & REM detection

Deep

33%

REM

49%

Sleep / wake & timing

2-state

87%

TST bias

−0.8 min

2-state: Vantage V · Miller 2022 · TST: Miller 2022

Manufacturer claim

No manufacturer 4-stage kappa is cited here. Polar's best wake-detection figures come from a Polar-sponsored study in children and teens (Pesonen 2018).

• How it tracks Wrist optical heart rate + motion; Polar sleep staging.
• Best for Total sleep time (−0.8 min)
• Weakest for Weak staging (0.28) and deep-sleep detection (33%)

View Polar on Amazon

Garmin Forerunner 245 Music

κ 0.25

What the studies found vs lab PSG

Kappa 0.25 in healthy young adults (Miller 2022, in-lab PSG), the lowest of the devices tested in healthy adults, with deep 28% and REM 50% caught. It catches only 27% of wake, so it over-reports total sleep by 43.8 min. The Vivosmart 4 scored 0.21 in a mixed sample where about half had suspected sleep apnea (Schyvens 2025).

4-stage staging (κ)

Healthy

0.25

Clinical

not tested

also 0.21 for the Vivosmart 4 in a mixed sample (Schyvens 2025)

Deep & REM detection

Deep

28%

REM

50%

Sleep / wake & timing

2-state

89%

TST bias

+43.8 min

2-state: Forerunner 245 · Miller 2022 · TST: Miller 2022

Manufacturer claim

No manufacturer 4-stage staging-accuracy figure is published. Garmin's sleep staging uses the Firstbeat engine; independent testing puts its 4-stage kappa at 0.25 in healthy adults.

• How it tracks Wrist optical heart rate + motion (Firstbeat).
• Best for Sleep/wake agreement (89%), the measure that runs high for every device
• Weakest for Weakest staging (0.25) and deep sleep (28%); over-reports sleep by 43.8 min

View Garmin on Amazon

Samsung Galaxy Watch 5

κ 0.42 clinical

What the studies found vs lab PSG

No healthy-adult PSG test. In adults with sleep complaints it scored kappa 0.42, catching 47.5% of deep sleep and 62.7% of REM (Lee 2023), with the best sleep-efficiency estimate of the 11 devices tested (−0.4 points).

4-stage staging (κ)

Healthy

not tested

Clinical

0.42

Deep & REM detection

Deep

n/a

REM

n/a

clinical: deep 47.5% · REM 62.7%

Sleep / wake & timing

2-state

n/a

TST bias

n/a

TST: no healthy-adult test

Manufacturer claim

No manufacturer 4-stage kappa is cited here. The only published 4-stage figure is the clinical one above (Lee 2023).

• How it tracks Wrist PPG + motion; Samsung Health sleep staging.
• Best for Sleep efficiency in sleep-complaint patients (−0.4 points)
• Weakest for No healthy-adult PSG test of any Galaxy Watch

View Samsung on Amazon

Fitbit Charge 5

κ 0.41 mixed

What the studies found vs lab PSG

Only tested in a mixed sample where about half had suspected sleep apnea (Schyvens 2025): kappa 0.41, deep 51.5%, REM 60.0%, total sleep +11.1 min. With no healthy-adult PSG figure, it is listed but not ranked.

4-stage staging (κ)

Healthy

not tested

Clinical

not tested

also 0.41 in a mixed sample (Schyvens 2025)

Sleep / wake & timing

2-state

n/a

TST bias

n/a

TST: no healthy-adult test

Manufacturer claim

Covered by the same Google/Fitbit algorithm update (kappa 0.63, 77% accuracy, 2026 in press) as the Sense line; independent validation still pending.

• How it tracks Wrist band; optical PPG + motion, Sleep Stages algorithm.
• Best for A low-cost band on the same Fitbit staging algorithm
• Weakest for No healthy-adult PSG test; shares the wake-detection weakness

View Fitbit on Amazon

Withings ScanWatch

κ 0.22 mixed

What the studies found vs lab PSG

Only tested in a mixed sample where about half had suspected sleep apnea (Schyvens 2025): kappa 0.22 on 3 stages (it merges deep and REM), with total sleep +39.9 min, the largest over-report in that study. With no healthy-adult PSG figure, it is listed but not ranked.

4-stage staging (κ)

Healthy

not tested

Clinical

not tested

also 0.22 on 3 stages in a mixed sample (Schyvens 2025)

Sleep / wake & timing

2-state

n/a

TST bias

n/a

TST: no healthy-adult test

Manufacturer claim

No manufacturer 4-stage staging-accuracy figure is published. The ScanWatch's validated strengths are ECG / afib and SpO2 detection, not sleep staging.

• How it tracks Hybrid analog watch; wrist PPG + accelerometer.
• Best for ECG and clinical-grade heart features (not sleep staging)
• Weakest for No healthy-adult PSG test; largest total-sleep over-report in its study (+39.9 min)

The "View on Amazon" links above are affiliate links. As an Amazon Associate, Kygo Health earns from qualifying purchases, at no extra cost to you.

True for every device

What no brand can fix.

Five findings that hold across every wearable in the tables. Read these before you trust any single number.

Universal

No device reliably measures wake

Every wearable catches most of your sleep but misses a large share of wake: from 27% of wake caught (Garmin) to 73% (Oura) in healthy adults. Most overestimate sleep and underestimate time awake at night, by up to 48 min. No wrist or ring reliably tells you how long you were actually awake.

Robbins 2024 · Svensson 2024 · Miller 2022

Age effect

Accuracy drops sharply with age

In older adults, total sleep time is underestimated by roughly 75 min and deep-sleep error explodes (MAPE 400–700%) on both the Fitbit Sense 2 and the Oura Gen 3. Numbers from young cohorts do not generalize.

Searles 2026

Better tier

EEG headbands beat every ring & watch

Muse-S and ear-EEG headbands reach kappa ~0.76, above any consumer ring or watch and matching human scorers. If you need genuine staging accuracy, that is the class to look at, not a wrist device.

npj Biomedical Innovations 2025 meta

Form factor

The ring leads vitals and stages

Oura leads the vitals measured during sleep: HRV (CCC 0.99), resting HR (CCC 0.98), temperature. It also leads 4-stage staging (kappa 0.65 vs Apple Watch 0.60 in the same study). Apple Watch's edge is REM detection (82.6%). Both share the same wake-detection weakness.

Dial 2025 · Robbins 2024

Coverage gap

The newest models are unvalidated

Every independent staging study tested a prior generation. Oura Ring 4/5, WHOOP 5/MG, Apple S9/S10/Ultra, Galaxy Watch/Ring and Pixel carry manufacturer-only claims, because peer review lags hardware by 18–36 months.

Validation coverage gap, 2026

The bottom line

Oura leads sleep staging (κ 0.65 in healthy adults, Oura-funded Robbins 2024) and total sleep time (about 2 min off). Apple Watch is second (κ 0.60) with the best REM detection (82.6%), and Fitbit third (κ 0.55) while holding up best in sleep-clinic patients (0.42). WHOOP (0.44), Polar (0.28) and Garmin (0.25) trail on staging, and Garmin over-reports sleep the most (+43.8 min). But every device overestimates sleep, can't reliably measure wake, and collapses in older adults and clinical patients, so use the trend, not the absolute number.

Keep exploring

Related tools.

More free, evidence-based tools to get the most out of your wearable.

Metrics tracked

Oura

Garmin

Apple

Fitbit

Sleep Metrics Comparison

Compare 38 sleep metrics tracked by Oura, Fitbit, Apple Watch and Garmin across 10 categories.

Sleep · 29 sources

Open

What moves accuracy

Wearable Accuracy: What Actually Affects It

Tattoos, hairy arms, cold hands, strap tightness, which wrist. What the studies found and the free fixes that work tonight.

Wearables · 26 answers

Open

Sleep stages overnight

Deep Sleep Factor Explorer

28 factors that affect deep sleep, ranked by research with doses and mechanisms.

Sleep · 28 factors

Open

Want every brand on every metric in one place? Open the wearable accuracy hub

FAQ

Common questions.

Which wearable is the most accurate for sleep?

Oura leads on sleep staging: Cohen's kappa 0.65 in healthy adults, ahead of Apple Watch at 0.60 and Fitbit at 0.55 in the same in-lab study (Robbins 2024, Oura-funded). Oura is also closest on total sleep time (about 2 minutes off). WHOOP (0.44), Polar (0.28) and Garmin (0.25) trail. No consumer device is "accurate" in an absolute sense: two human sleep scorers agree at about 0.75.

Are sleep trackers accurate for sleep stages?

Only moderately. In healthy adults the best consumer wearables reach a 4-stage kappa of about 0.55 to 0.65, against about 0.75 between two human scorers. Every device overestimates sleep, misses wake, and drops sharply in older adults and in people with sleep disorders. Use the trend over weeks, not a single night's stage breakdown.

Why do the accuracy numbers drop for people with sleep problems?

It is a population effect, not a study disagreement. Screened healthy sleepers score far higher than real sleep-clinic patients. Oura falls from kappa 0.65 (healthy) to 0.35 (clinical); Apple from 0.60 to 0.30. Apnea-fragmented sleep is simply harder to stage, so every device drops. Fitbit holds up best, from 0.55 to 0.42.

Can my watch or ring tell how long I was awake?

Not reliably. Every device catches most of your sleep but misses a large share of wake: from 27% of wake caught (Garmin) to 73% (Oura) in healthy adults. Most overestimate sleep and underestimate time awake at night, by up to 48 minutes. Wake detection is the universal weakness across every brand.

Do rings or watches track sleep better?

In the studies so far, the ring leads. Oura leads the vitals measured during sleep (heart-rate variability, resting heart rate and temperature), 4-stage staging (kappa 0.65 vs Apple Watch 0.60 in the same study) and total sleep time. Apple Watch has the best REM detection (82.6%). Both share the wake-detection weakness.

Are the newest models like Oura Ring 4, WHOOP 5 and Apple Series 10 more accurate?

Unknown independently. Every published validation study tested a prior generation, so all current-generation accuracy claims are manufacturer-only. Peer-reviewed validation lags new hardware by roughly 18 to 36 months, so newer does not yet mean proven.

Sources

Every claim, traceable.

Each figure checked against the primary record (PubMed / journal / manufacturer paper). Verified October 2026.

Independent · mixed sample

Schyvens et al. 2025: Six wrist-worn sleep trackers vs PSG

Sleep Advances. 2025;6(2):zpaf021. N=62, mixed sample: about half with suspected sleep apnea.

Independent · clinical

Lee T et al. 2023: 11 devices in sleep-clinic patients

JMIR mHealth uHealth. 2023;11:e50983. N=75, mean AHI 18.

Independent · clinical

Herberger et al. 2025: Oura Gen 3 + 2 rings vs PSG

Scientific Reports. 2025;15:9461. N=45 sleep-clinic patients.

Independent · meta

Khan et al. 2025: Oura sleep meta-analysis

OTO Open. 2025;9:e70181. 6 studies, N=388 (COI: none).

Independent · meta

JCSM wrist-wearable sleep meta (Lee et al.)

J Clin Sleep Med. 24 studies, N=798.

Healthy · WHOOP-funded lab

Miller et al. 2022: Six devices vs PSG

Sensors. 2022;22(16):6317. N=53 healthy young adults. WHOOP 3.0 0.44, Polar Vantage V 0.28, Garmin Forerunner 245 0.25.

Independent · Fitbit meta

Haghayegh et al. 2019: Fitbit sleep-stage meta

JMIR. 2019;21(11):e16273. PMC6908975 (COI: none).

Independent · age effect

Searles et al. 2026: Age & sleep-tracker accuracy

Sleep Advances. 2026;7(1):zpag006. Fitbit Sense 2 + Oura Gen 3.

Independent · EEG tier

npj Biomedical Innovations 2025: Wearable-EEG meta

43 studies. Muse-S / ear-EEG kappa ~0.76.

Independent · vitals

Dial et al. 2025: Oura Gen 4 HRV / RHR vs ECG

Physiological Reports. 2025;13:e70527 (AFRL).

Oura-funded · healthy

Robbins et al. 2024: Oura Gen 3, Apple Watch, Fitbit vs PSG

Sensors. 2024;24(20):6532. Healthy adults; Oura 0.65, Apple Watch S8 0.60, Fitbit Sense 2 0.55.

Oura-funded · healthy

Svensson et al. 2024: Oura Gen 3 (OSSA 2.0) vs home PSG

Sleep Medicine. 2024. 96 healthy adults, multi-night; deep 64.0%, REM 72.6%, total sleep −2.1 min.

Manufacturer

Apple Watch sleep-staging white paper (Sept 2023)

~1,400 nights; kappa 0.63; sens 97.8% / spec 76%.

Manufacturer · in press

Google/Fitbit 2026: Updated staging algorithm

kappa 0.63 (up from 0.56); accuracy 77% (up from 71%).

Show all 14 sources

From the blog

Show fewer sources

Keep reading.

The long-form, evidence-based articles behind this tool.

Sleep

Most Accurate Sleep Tracker? Oura vs Apple vs Fitbit vs Garmin vs Whoop (2026)

Every tracker calls itself accurate. Measured against lab polysomnography, the ranking changes completely on one thing most reviews ignore.

8 min read

Read

Sleep

Every Sleep Metric Your Wearable Tracks: Oura vs Apple Watch vs Fitbit vs Garmin Compared

What Oura, Apple Watch, Fitbit and Garmin each actually measure overnight, what they miss, and why the gaps matter.

9 min read

Read

Sleep

Why Is My Sleep Score Low When I Slept 8 Hours?

Your tracker says eight hours and your score is still bad. Why duration is not quality, and what a food log reveals about the cause.

8 min read

Read

This content is for informational purposes only and is not medical advice. Wearable sleep staging is an estimate, not a clinical measurement. Consult a qualified healthcare provider before making health decisions based on wearable data.

Data from peer-reviewed validation studies and manufacturer documentation. Last updated October 25, 2026.

As an Amazon Associate, Kygo Health earns from qualifying purchases. Product links on this page are affiliate links; we may earn a commission at no extra cost to you.