Device comparisons
Whoop vs Apple Watch: What the Sleep-Lab Studies Found
Two independent sleep-lab studies tested WHOOP and Apple Watch on the same people on the same night, against polysomnography (PSG) for sleep and ECG for heart rate. Miller and colleagues (2022) tested WHOOP 3.0 and Apple Watch Series 6 in 53 adults; Schyvens and colleagues (2025) tested WHOOP 4.0 and Apple Watch Series 8 in 62 adults. Overnight heart rate was close to ECG for both. WHOOP 3.0 tracked overnight HRV more closely than Apple Watch S6. For sleep stages the two studies point in different directions, so neither device clearly wins there.
Reviewed by Sameer Sontakey, Co-founder and Chief AI & Algorithms Officer at Sensor Bio. ChatPPG is independent and device-neutral; it is not a Sensor Bio product.
Specs verified on October 16, 2026: Miller, Sargent & Roach 2022, Sensors (WHOOP 3.0, Apple Watch S6, n=53), Schyvens et al. 2025, Sleep Advances (WHOOP 4.0, Apple Watch S8, n=62).
Specifications Comparison
| Specification | WHOOP | Apple Watch |
|---|---|---|
| Models tested | WHOOP 3.0 (Miller 2022), WHOOP 4.0 (Schyvens 2025) | Series 6 (Miller 2022), Series 8 (Schyvens 2025) |
| Overnight HR vs ECG (Miller) | Bias −0.3 BPM, absolute 0.7 BPM, ICC 0.99 | Bias +0.5 BPM, absolute 1.5 BPM, ICC 0.96 |
| Overnight HRV vs ECG (Miller) | Bias −4.5 ms, absolute 4.7 ms, ICC 0.99 | Bias −9.6 ms, absolute 22.5 ms, ICC 0.67 |
| Sleep vs wake (Miller) | 86% of epochs, kappa 0.44 | 88% of epochs, kappa 0.30 |
| Sleep stages (Miller) | 60% of epochs, kappa 0.44 | 53% of epochs, kappa 0.20 |
| Sleep stages (Schyvens) | Kappa 0.37 (n=40) | Kappa 0.53 (n=20) |
Pros & Cons
WHOOP
Pros
- + Overnight HRV almost matched ECG in Miller 2022 (WHOOP 3.0: mean bias −4.5 ms, ICC 0.99)
- + Overnight heart rate within 0.7 BPM of ECG on average (WHOOP 3.0, ICC 0.99)
- + Best of six devices at spotting deep sleep in Schyvens 2025 (WHOOP 4.0 matched 69.6% of PSG N3 epochs)
Cons
- - Overestimated deep sleep by 31.5 min and REM by 15.3 min on average in Schyvens 2025 (WHOOP 4.0)
- - Lower whole-night stage agreement than Apple Watch S8 in Schyvens 2025 (kappa 0.37 vs 0.53)
- - Screenless band: no on-wrist display of live heart rate
Apple Watch
Pros
- + Overnight heart rate close to ECG (Series 6: mean bias +0.5 BPM, ICC 0.96)
- + Highest stage agreement of six devices in Schyvens 2025 (Series 8, kappa 0.53)
- + Best at spotting wake (52.2%) and REM (68.6%) epochs in Schyvens 2025
Cons
- - Overestimated total sleep time by 39.5 min on average in Miller 2022 (Series 6)
- - Overnight HRV agreement only good, not excellent, in Miller 2022 (bias −9.6 ms, absolute 22.5 ms, ICC 0.67)
- - Underestimated deep sleep by 25.2 min in Schyvens 2025 (Series 8)
What published studies show
Results quoted from 37 published validation studies, each against a clinical reference. A result applies to the model and protocol tested, not to every model the brand sells. From the wearable accuracy database (v1.0.0).
Heart rate at rest3 studies
Only maker-funded evidence for WHOOP on this metric.
| Device tested | Result | Reference | Sample | Study |
|---|---|---|---|---|
| WHOOP | ≤ 0.39 ± 0.38%Mean bias (signed) | ECG | Rest, 15 opportunities | Bellenger et al. (2021)Maker funded |
| Apple Watch Series 9 / Ultra 2 | 5.91% (MAE 3.73 bpm)Mean absolute percentage error | Polar H10 + Kubios HRV | n = 39, Serial morning measurements over 14 days, 316 readings | O'Grady et al. (2024)Independent |
| Apple Watch Series 6R-R intervals and BPM at rest | 1.15%Mean absolute percentage error | ECG | n = 78, Lab: rest, talking, movie, before/after walking | Bonneval et al. (2025)Independent |
- WHOOP≤ 0.39 ± 0.38%
Mean bias (signed) · vs ECG · Rest, 15 opportunities
Bellenger et al. (2021) Maker funded
- Apple Watch Series 9 / Ultra 25.91% (MAE 3.73 bpm)
Mean absolute percentage error · vs Polar H10 + Kubios HRV · n = 39, Serial morning measurements over 14 days, 316 readings
O'Grady et al. (2024) Independent
- Apple Watch Series 61.15%
Mean absolute percentage error, R-R intervals and BPM at rest · vs ECG · n = 78, Lab: rest, talking, movie, before/after walking
Bonneval et al. (2025) Independent
Heart rate during exercise10 studies
| Device tested | Result | Reference | Sample | Study |
|---|---|---|---|---|
| WHOOP 3.0 | High agreement only in squats; low in rows and burpeesStudy finding (no single number) | ECG chest strap (Polar H7/H10) | n = 29, Resistance exercise (5 exercises) | Støve et al. (2022)Funding not reported |
| WHOOP 4.0 (upper arm) | Strongest agreement (with Verity Sense) in graded treadmillStudy finding (no single number) | ECG chest strap (Polar H7/H10) | n = 28, Rest, warm-up, burpees, Bruce treadmill | Moghaddam et al. (2025)Funding not reported |
| Apple Watch Series 6 | High agreement in squat, deadlift, cable row (rho > 0.832)Study finding (no single number) | ECG chest strap (Polar H7/H10) | n = 29, Resistance exercise (5 exercises) | Støve et al. (2022)Funding not reported |
| Apple Watch SE | Significantly lower MAE than Polar Pacer Pro and Galaxy Watch 6Study finding (no single number) | ECG chest strap (Polar H7/H10) | n = 37, Low/moderate/high-intensity treadmill | Jeong et al. (2026)Independent |
| Apple Watch Series 8 | Skin pigmentation did not predict error (P = 0.62); 50% of missing data from dark-skin participantsStudy finding (no single number) | ECG chest strap (Polar H7/H10) | n = 28, Rest, cycling, recovery | Mulholland et al. (2026)Independent |
| Apple Watch Series 9 | Less accurate for darker skin pigmentationStudy finding (no single number) | ECG chest strap (Polar H7/H10) | n = 30, 10-min treadmill | Chase et al. (2024)Independent |
- WHOOP 3.0High agreement only in squats; low in rows and burpees
Study finding (no single number) · vs ECG chest strap (Polar H7/H10) · n = 29, Resistance exercise (5 exercises)
Støve et al. (2022) Funding not reported
- WHOOP 4.0 (upper arm)Strongest agreement (with Verity Sense) in graded treadmill
Study finding (no single number) · vs ECG chest strap (Polar H7/H10) · n = 28, Rest, warm-up, burpees, Bruce treadmill
Moghaddam et al. (2025) Funding not reported
- Apple Watch Series 6High agreement in squat, deadlift, cable row (rho > 0.832)
Study finding (no single number) · vs ECG chest strap (Polar H7/H10) · n = 29, Resistance exercise (5 exercises)
Støve et al. (2022) Funding not reported
- Apple Watch SESignificantly lower MAE than Polar Pacer Pro and Galaxy Watch 6
Study finding (no single number) · vs ECG chest strap (Polar H7/H10) · n = 37, Low/moderate/high-intensity treadmill
Jeong et al. (2026) Independent
- Apple Watch Series 8Skin pigmentation did not predict error (P = 0.62); 50% of missing data from dark-skin participants
Study finding (no single number) · vs ECG chest strap (Polar H7/H10) · n = 28, Rest, cycling, recovery
Mulholland et al. (2026) Independent
- Apple Watch Series 9Less accurate for darker skin pigmentation
Study finding (no single number) · vs ECG chest strap (Polar H7/H10) · n = 30, 10-min treadmill
Chase et al. (2024) Independent
5 more results in the database. Heart rate accuracy across devices →
Overnight heart rate2 studies
| Device tested | Result | Reference | Sample | Study |
|---|---|---|---|---|
| WHOOP 4.0 | 3.00%Mean absolute percentage error | ECG | n = 13, Home, during sleep, 536 nights | Dial et al. (2025)Independent |
| WHOOP 4.0 | 0.91Lin's concordance correlation | ECG | n = 13, Home, during sleep, 536 nights | Dial et al. (2025)Independent |
| WHOOP 3.0 | ICC 0.99Intraclass correlation | ECG | n = 53, Overnight, sleep laboratory | Miller et al. (2022)Maker funded |
| WHOOP 3.0 | 0.7 bpm (absolute bias)Mean absolute error | ECG | n = 53, Overnight, sleep laboratory | Miller et al. (2022)Maker funded |
| Apple Watch Series 6 | ICC 0.96Intraclass correlation | ECG | n = 53, Overnight, sleep laboratory | Miller et al. (2022)Other maker funded |
| Apple Watch Series 6 | 1.5 bpm (absolute bias)Mean absolute error | ECG | n = 53, Overnight, sleep laboratory | Miller et al. (2022)Other maker funded |
- WHOOP 4.03.00%
Mean absolute percentage error · vs ECG · n = 13, Home, during sleep, 536 nights
Dial et al. (2025) Independent
- WHOOP 4.00.91
Lin's concordance correlation · vs ECG · n = 13, Home, during sleep, 536 nights
Dial et al. (2025) Independent
- WHOOP 3.0ICC 0.99
Intraclass correlation · vs ECG · n = 53, Overnight, sleep laboratory
Miller et al. (2022) Maker funded
- WHOOP 3.00.7 bpm (absolute bias)
Mean absolute error · vs ECG · n = 53, Overnight, sleep laboratory
Miller et al. (2022) Maker funded
- Apple Watch Series 6ICC 0.96
Intraclass correlation · vs ECG · n = 53, Overnight, sleep laboratory
Miller et al. (2022) Other maker funded
- Apple Watch Series 61.5 bpm (absolute bias)
Mean absolute error · vs ECG · n = 53, Overnight, sleep laboratory
Miller et al. (2022) Other maker funded
HRV (RMSSD)4 studies
| Device tested | Result | Reference | Sample | Study |
|---|---|---|---|---|
| WHOOP 4.0 | 8.17%Mean absolute percentage error | ECG | n = 13, Home, during sleep, 536 nights | Dial et al. (2025)Independent |
| WHOOP 4.0 | 0.94Lin's concordance correlation | ECG | n = 13, Home, during sleep, 536 nights | Dial et al. (2025)Independent |
| WHOOP | Ln RMSSD 1.66 ± 1.80% (LoA ±5.93%)Mean bias (signed) | ECG | Rest, 15 opportunities | Bellenger et al. (2021)Maker funded |
| WHOOP 3.0 | ICC 0.99Intraclass correlation | ECG | n = 53, Overnight, sleep laboratory | Miller et al. (2022)Maker funded |
| WHOOP 3.0 | 4.7 ms (absolute bias)Mean absolute error (absolute bias) | ECG | n = 53, Overnight, sleep laboratory | Miller et al. (2022)Maker funded |
| Apple Watch Series 9 / Ultra 2 | 28.88% (MAE 20.46 ms)Mean absolute percentage error | Polar H10 + Kubios HRV | n = 39, Serial morning measurements over 14 days, 316 readings | O'Grady et al. (2024)Independent |
- WHOOP 4.08.17%
Mean absolute percentage error · vs ECG · n = 13, Home, during sleep, 536 nights
Dial et al. (2025) Independent
- WHOOP 4.00.94
Lin's concordance correlation · vs ECG · n = 13, Home, during sleep, 536 nights
Dial et al. (2025) Independent
- WHOOPLn RMSSD 1.66 ± 1.80% (LoA ±5.93%)
Mean bias (signed) · vs ECG · Rest, 15 opportunities
Bellenger et al. (2021) Maker funded
- WHOOP 3.0ICC 0.99
Intraclass correlation · vs ECG · n = 53, Overnight, sleep laboratory
Miller et al. (2022) Maker funded
- WHOOP 3.04.7 ms (absolute bias)
Mean absolute error (absolute bias) · vs ECG · n = 53, Overnight, sleep laboratory
Miller et al. (2022) Maker funded
- Apple Watch Series 9 / Ultra 228.88% (MAE 20.46 ms)
Mean absolute percentage error · vs Polar H10 + Kubios HRV · n = 39, Serial morning measurements over 14 days, 316 readings
O'Grady et al. (2024) Independent
2 more results in the database. HRV accuracy across devices →
Sleep stages (4-stage)6 studies
| Device tested | Result | Reference | Sample | Study |
|---|---|---|---|---|
| WHOOP (strap, 2.0 era) | 64% (κ 0.47)Epoch-by-epoch agreement | Polysomnography (PSG) | n = 12, Sleep laboratory, 10-day protocol, 86 sleeps | Miller et al. (2020)Funding not reported |
| WHOOP (strap, 2.0 era) | κ = 0.47Cohen's kappa | Polysomnography (PSG) | n = 12, Sleep laboratory, 10-day protocol, 86 sleeps | Miller et al. (2020)Funding not reported |
| WHOOP (auto-detect) | 63%Epoch-by-epoch agreement | Polysomnography (PSG) | n = 6, Sleep laboratory, 9 nights, 54 sleeps | Miller et al. (2021)Maker funded |
| WHOOP 3.0 | 60%Epoch-by-epoch agreement | Polysomnography (PSG) | n = 53, Sleep laboratory, 1 night, 9 h in bed, all devices worn together | Miller et al. (2022)Maker funded |
| WHOOP 3.0 | κ = 0.44Cohen's kappa | Polysomnography (PSG) | n = 53, Sleep laboratory, 1 night, 9 h in bed, all devices worn together | Miller et al. (2022)Maker funded |
| WHOOP 4.0 | κ = 0.37Cohen's kappa | Polysomnography (PSG) | n = 40, Sleep laboratory, 1 night | Schyvens et al. (2025)Funding not reported |
- WHOOP (strap, 2.0 era)64% (κ 0.47)
Epoch-by-epoch agreement · vs Polysomnography (PSG) · n = 12, Sleep laboratory, 10-day protocol, 86 sleeps
Miller et al. (2020) Funding not reported
- WHOOP (strap, 2.0 era)κ = 0.47
Cohen's kappa · vs Polysomnography (PSG) · n = 12, Sleep laboratory, 10-day protocol, 86 sleeps
Miller et al. (2020) Funding not reported
- WHOOP (auto-detect)63%
Epoch-by-epoch agreement · vs Polysomnography (PSG) · n = 6, Sleep laboratory, 9 nights, 54 sleeps
Miller et al. (2021) Maker funded
- WHOOP 3.060%
Epoch-by-epoch agreement · vs Polysomnography (PSG) · n = 53, Sleep laboratory, 1 night, 9 h in bed, all devices worn together
Miller et al. (2022) Maker funded
- WHOOP 3.0κ = 0.44
Cohen's kappa · vs Polysomnography (PSG) · n = 53, Sleep laboratory, 1 night, 9 h in bed, all devices worn together
Miller et al. (2022) Maker funded
- WHOOP 4.0κ = 0.37
Cohen's kappa · vs Polysomnography (PSG) · n = 40, Sleep laboratory, 1 night
Schyvens et al. (2025) Funding not reported
7 more results in the database. Sleep accuracy across devices →
Sleep vs wake (2-stage)5 studies
| Device tested | Result | Reference | Sample | Study |
|---|---|---|---|---|
| WHOOP 4.0 | > 90% (all devices)Sensitivity | Polysomnography (PSG) | n = 62, Sleep laboratory, 1 night | Schyvens et al. (2025)Funding not reported |
| WHOOP (strap, 2.0 era) | 89% (κ 0.49)Epoch-by-epoch agreement | Polysomnography (PSG) | n = 12, Sleep laboratory, 10-day protocol, 86 sleeps | Miller et al. (2020)Funding not reported |
| WHOOP (strap, 2.0 era) | κ = 0.49Cohen's kappa | Polysomnography (PSG) | n = 12, Sleep laboratory, 10-day protocol, 86 sleeps | Miller et al. (2020)Funding not reported |
| WHOOP (auto-detect) | 86%Epoch-by-epoch agreement | Polysomnography (PSG) | n = 6, Sleep laboratory, 9 nights, 54 sleeps | Miller et al. (2021)Maker funded |
| WHOOP 3.0 | 86%Epoch-by-epoch agreement | Polysomnography (PSG) | n = 53, Sleep laboratory, 1 night, 9 h in bed, all devices worn together | Miller et al. (2022)Maker funded |
| WHOOP 3.0 | κ = 0.44Cohen's kappa | Polysomnography (PSG) | n = 53, Sleep laboratory, 1 night, 9 h in bed, all devices worn together | Miller et al. (2022)Maker funded |
- WHOOP 4.0> 90% (all devices)
Sensitivity · vs Polysomnography (PSG) · n = 62, Sleep laboratory, 1 night
Schyvens et al. (2025) Funding not reported
- WHOOP (strap, 2.0 era)89% (κ 0.49)
Epoch-by-epoch agreement · vs Polysomnography (PSG) · n = 12, Sleep laboratory, 10-day protocol, 86 sleeps
Miller et al. (2020) Funding not reported
- WHOOP (strap, 2.0 era)κ = 0.49
Cohen's kappa · vs Polysomnography (PSG) · n = 12, Sleep laboratory, 10-day protocol, 86 sleeps
Miller et al. (2020) Funding not reported
- WHOOP (auto-detect)86%
Epoch-by-epoch agreement · vs Polysomnography (PSG) · n = 6, Sleep laboratory, 9 nights, 54 sleeps
Miller et al. (2021) Maker funded
- WHOOP 3.086%
Epoch-by-epoch agreement · vs Polysomnography (PSG) · n = 53, Sleep laboratory, 1 night, 9 h in bed, all devices worn together
Miller et al. (2022) Maker funded
- WHOOP 3.0κ = 0.44
Cohen's kappa · vs Polysomnography (PSG) · n = 53, Sleep laboratory, 1 night, 9 h in bed, all devices worn together
Miller et al. (2022) Maker funded
4 more results in the database. Sleep accuracy across devices →
Total sleep time1 study
| Device tested | Result | Reference | Sample | Study |
|---|---|---|---|---|
| WHOOP (strap, 2.0 era) | +8.2 ± 32.9 min (n.s.)Mean bias (signed) | Polysomnography (PSG) | n = 12, Sleep laboratory, 10-day protocol, 86 sleeps | Miller et al. (2020)Funding not reported |
- WHOOP (strap, 2.0 era)+8.2 ± 32.9 min (n.s.)
Mean bias (signed) · vs Polysomnography (PSG) · n = 12, Sleep laboratory, 10-day protocol, 86 sleeps
Miller et al. (2020) Funding not reported
Blood oxygen (SpO2)8 studies
| Device tested | Result | Reference | Sample | Study |
|---|---|---|---|---|
| Apple Watch Series 7 | 3.5% (rest)Root-mean-square error | Medical-grade pulse oximeter | n = 36, Rest, COPD outpatients | Støve et al. (2023)Funding not reported |
| Apple Watch Series 6 | r = 0.89Pearson correlation | Medical-grade pulse oximeter | n = 265, Spot readings | Spaccarotella et al. (2022)Independent |
| Apple Watch Series 6 | −0.23 (LoA −3.49 to 3.04)Mean bias (signed) | Medical-grade pulse oximeter | n = 265, Spot readings | Spaccarotella et al. (2022)Independent |
| Apple Watch Series 7 | 2.2% (RMSE 2.9%)Mean absolute error | Medical-grade pulse oximeter | n = 49, Spot readings, with and without chronic pulmonary disease | Jiang et al. (2023)Independent |
| Apple Watch Series 7 | 2.9%Root-mean-square error | Medical-grade pulse oximeter | n = 49, Spot readings, with and without chronic pulmonary disease | Jiang et al. (2023)Independent |
| Apple Watch Series 8 | Met ISO RMS ≤ 4% at normoxia and SpO2 < 90%; lowest bias and RMSD of the threeStudy finding (no single number) | Medical-grade pulse oximeter | n = 18, Induced hypoxemia (FiO2 14%, 12%, 10%), ISO 80601-2-61 | Walzel et al. (2023)Independent |
- Apple Watch Series 73.5% (rest)
Root-mean-square error · vs Medical-grade pulse oximeter · n = 36, Rest, COPD outpatients
Støve et al. (2023) Funding not reported
- Apple Watch Series 6r = 0.89
Pearson correlation · vs Medical-grade pulse oximeter · n = 265, Spot readings
Spaccarotella et al. (2022) Independent
- Apple Watch Series 6−0.23 (LoA −3.49 to 3.04)
Mean bias (signed) · vs Medical-grade pulse oximeter · n = 265, Spot readings
Spaccarotella et al. (2022) Independent
- Apple Watch Series 72.2% (RMSE 2.9%)
Mean absolute error · vs Medical-grade pulse oximeter · n = 49, Spot readings, with and without chronic pulmonary disease
Jiang et al. (2023) Independent
Root-mean-square error · vs Medical-grade pulse oximeter · n = 49, Spot readings, with and without chronic pulmonary disease
Jiang et al. (2023) Independent
- Apple Watch Series 8Met ISO RMS ≤ 4% at normoxia and SpO2 < 90%; lowest bias and RMSD of the three
Study finding (no single number) · vs Medical-grade pulse oximeter · n = 18, Induced hypoxemia (FiO2 14%, 12%, 10%), ISO 80601-2-61
Walzel et al. (2023) Independent
4 more results in the database. SpO2 accuracy across devices →
Atrial fibrillation detection4 studies
| Device tested | Result | Reference | Sample | Study |
|---|---|---|---|---|
| Apple Watch Series 6smartwatch ECG app, inconclusive tracings counted | 85% (spec 75%)Sensitivity | 12-lead ECG | n = 201, Tertiary cardiology service, single-lead ECG algorithms | Mannhart et al. (2023)Funding not reported |
| Apple Watch Series 6smartwatch ECG app, inconclusive tracings counted | 75%Specificity | 12-lead ECG | n = 201, Tertiary cardiology service, single-lead ECG algorithms | Mannhart et al. (2023)Funding not reported |
| Apple Watch + KardiaBand (SmartRhythm 2.0)AF episodes ≥ 1 h | 97.5% episode sensitivity (PPV 39.9%)Sensitivity | Insertable cardiac monitor | n = 24, Ambulatory, 31,348.9 h of recordings | Wasserlauf et al. (2019)Funding not reported |
| Apple Watch (irregular rhythm notification)PPG irregular pulse notification | PPV 0.84 (95% CI 0.76–0.92)Positive predictive value | ECG patch | n = 419297, Siteless pragmatic study; ECG patch after notification | Perez et al. (2019)Maker funded |
| Kardia Band on Apple Watch (single-lead ECG)automated ECG read; 57 of 169 recordings non-interpretable | Sens 93% / spec 84% (interpretable recordings)Sensitivity | 12-lead ECG | n = 100, Before and after cardioversion | Bumgarner et al. (2018)Funding not reported |
- Apple Watch Series 685% (spec 75%)
Sensitivity, smartwatch ECG app, inconclusive tracings counted · vs 12-lead ECG · n = 201, Tertiary cardiology service, single-lead ECG algorithms
Mannhart et al. (2023) Funding not reported
Specificity, smartwatch ECG app, inconclusive tracings counted · vs 12-lead ECG · n = 201, Tertiary cardiology service, single-lead ECG algorithms
Mannhart et al. (2023) Funding not reported
- Apple Watch + KardiaBand (SmartRhythm 2.0)97.5% episode sensitivity (PPV 39.9%)
Sensitivity, AF episodes ≥ 1 h · vs Insertable cardiac monitor · n = 24, Ambulatory, 31,348.9 h of recordings
Wasserlauf et al. (2019) Funding not reported
- Apple Watch (irregular rhythm notification)PPV 0.84 (95% CI 0.76–0.92)
Positive predictive value, PPG irregular pulse notification · vs ECG patch · n = 419297, Siteless pragmatic study; ECG patch after notification
Perez et al. (2019) Maker funded
- Kardia Band on Apple Watch (single-lead ECG)Sens 93% / spec 84% (interpretable recordings)
Sensitivity, automated ECG read; 57 of 169 recordings non-interpretable · vs 12-lead ECG · n = 100, Before and after cardioversion
Bumgarner et al. (2018) Funding not reported
VO2max estimate3 studies
| Device tested | Result | Reference | Sample | Study |
|---|---|---|---|---|
| Apple Watch Series 10 | 13.2% (MAE 6.79 mL/kg/min)Mean absolute percentage error | Indirect calorimetry | n = 40, Maximal treadmill test (modified Åstrand) | Lambe et al. (2026)Independent |
| Apple Watch | 13.31% (MAE 6.92 mL/kg/min)Mean absolute percentage error | Indirect calorimetry | n = 30, Maximal treadmill test (modified Åstrand) | Lambe et al. (2025)Independent |
| Apple Watch Series 7 | 15.79% (ICC 0.47)Mean absolute percentage error | Indirect calorimetry | n = 19, Graded cycle ergometer test | Caserman et al. (2024)Independent |
- Apple Watch Series 1013.2% (MAE 6.79 mL/kg/min)
Mean absolute percentage error · vs Indirect calorimetry · n = 40, Maximal treadmill test (modified Åstrand)
Lambe et al. (2026) Independent
- Apple Watch13.31% (MAE 6.92 mL/kg/min)
Mean absolute percentage error · vs Indirect calorimetry · n = 30, Maximal treadmill test (modified Åstrand)
Lambe et al. (2025) Independent
- Apple Watch Series 715.79% (ICC 0.47)
Mean absolute percentage error · vs Indirect calorimetry · n = 19, Graded cycle ergometer test
Caserman et al. (2024) Independent
Energy expenditure (calories)6 studies
| Device tested | Result | Reference | Sample | Study |
|---|---|---|---|---|
| Apple Watch Series 9 | Inconsistent across all groupsStudy finding (no single number) | Indirect calorimetry | n = 30, 10-min treadmill | Chase et al. (2024)Independent |
| Apple Watch Series 4 | 13.5%–27.1% (range by intensity)Coefficient of variation | Indirect calorimetry | n = 25, Sitting, walking, running, sprints | Düking et al. (2020)Independent |
| Apple Watch Series 6 | 14.68–24.85% (poor in all activities)Coefficient of variation | Indirect calorimetry | n = 60, Sitting, walking, running, resistance, cycling | Hajj-Boutros et al. (2023)Funding not reported |
| Apple Watch Series 8physical activity energy expenditure, 10 min | +21.6 kcalMean bias (signed) | Indirect calorimetry | n = 52, 10-min recumbent cycling intervals | Kostrna et al. (2026)Independent |
| Apple Watch Series 6walking | 19.8% walk / 24.4% runMean absolute percentage error | Indirect calorimetry | n = 20, Outdoor 2-km walk (6 km/h) and run (10 km/h) | Le et al. (2022)Independent |
| Apple Watch | Lowest EE error of 7 devices, but no device below 20%Study finding (no single number) | Indirect calorimetry | n = 60, Sitting, walking, running, cycling | Shcherbina et al. (2017)Independent |
- Apple Watch Series 9Inconsistent across all groups
Study finding (no single number) · vs Indirect calorimetry · n = 30, 10-min treadmill
Chase et al. (2024) Independent
- Apple Watch Series 413.5%–27.1% (range by intensity)
Coefficient of variation · vs Indirect calorimetry · n = 25, Sitting, walking, running, sprints
Düking et al. (2020) Independent
- Apple Watch Series 614.68–24.85% (poor in all activities)
Coefficient of variation · vs Indirect calorimetry · n = 60, Sitting, walking, running, resistance, cycling
Hajj-Boutros et al. (2023) Funding not reported
- Apple Watch Series 8+21.6 kcal
Mean bias (signed), physical activity energy expenditure, 10 min · vs Indirect calorimetry · n = 52, 10-min recumbent cycling intervals
Kostrna et al. (2026) Independent
- Apple Watch Series 619.8% walk / 24.4% run
Mean absolute percentage error, walking · vs Indirect calorimetry · n = 20, Outdoor 2-km walk (6 km/h) and run (10 km/h)
Le et al. (2022) Independent
- Apple WatchLowest EE error of 7 devices, but no device below 20%
Study finding (no single number) · vs Indirect calorimetry · n = 60, Sitting, walking, running, cycling
Shcherbina et al. (2017) Independent
Verdict
For overnight HRV, the head-to-head evidence favours WHOOP: in Miller 2022, WHOOP 3.0 matched ECG almost exactly while Apple Watch S6 was further off. For overnight heart rate, both were within about 1.5 BPM of ECG. For sleep stages, Miller found WHOOP 3.0 ahead of Apple Watch S6 (kappa 0.44 vs 0.20), while Schyvens found Apple Watch S8 ahead of WHOOP 4.0 (0.53 vs 0.37), so the result depends on model generation and study. Both studies tested healthy adults for a single night, and neither looked at heart rate during exercise.
Frequently Asked Questions
Is WHOOP or Apple Watch more accurate?
For overnight HRV, WHOOP 3.0 was closer to ECG than Apple Watch S6 in a 53-person sleep-lab study (ICC 0.99 vs 0.67). Overnight heart rate was close for both. For sleep stages the studies disagree: WHOOP 3.0 beat Apple Watch S6 in Miller 2022, but Apple Watch S8 beat WHOOP 4.0 in Schyvens 2025.
Which tracks sleep better, WHOOP or Apple Watch?
It depends on the model. In Schyvens 2025, Apple Watch Series 8 had the higher stage agreement with polysomnography (kappa 0.53 vs 0.37 for WHOOP 4.0), though WHOOP 4.0 was better at spotting deep sleep. In Miller 2022, WHOOP 3.0 beat Apple Watch S6 (0.44 vs 0.20).
Has anyone compared WHOOP and Apple Watch during exercise?
Not in the two head-to-head studies on this page; both were overnight sleep-lab studies. See our WHOOP and Apple Watch accuracy reviews for exercise studies of each device.
Related Research
Device research records
Datasets, papers and validation results for the models named on this page.