Sleep Tracker Accuracy: What Watches Get Right (and What They Invent)
Garmin says 1 h 40 m deep sleep; Oura says 48 minutes; you feel identical either way. That conflict is algorithm theater: consumer wearables estimate sleep from motion, HR, and HRV. They are not polysomnography (PSG), and stage labels are the weakest part of the report.
What Is Usually Good Enough
Total sleep time and bedtime consistency are the most useful fields for runners. Overnight HR and HRV trends can add context for recovery decisions when paired with how easy runs feel. Stage breakdowns (light/deep/REM) often disagree across brands and versus PSG — fine for curiosity, poor as a reason to cancel intervals. Related performance framing: sleep and running performance.
Worked Scenario: Two Devices, One Night
You wear a watch and a ring. Both agree you were in bed 11:00–6:30 and asleep roughly 6 h 50 m–7 h 10 m. Deep sleep: Device A 1:35, Device B 0:55. REM similarly mismatched. Training decision: use ~7 hours total and how you feel on the warm-up — ignore the deep-sleep argument. If total sleep was 5 h 20 m on both devices for three nights, that aligned duration signal is actionable even if stages disagree.
When Scores Mislead Training
- Charging gaps create “short sleep” artifacts
- Alcohol can fragment sleep while some scores look oddly “calm”
- Travel and late screens scramble timing without proving overtraining
- Firmware updates can shift baselines overnight
If two brands disagree, pick the one you will wear consistently and stop cross-checking stage charts. Coaching needs a stable baseline more than a winner algorithm.
Validation studies generally find wearable total sleep time closer to PSG than stage classification — but “closer” is not “equal.” Proprietary scores (readiness, body battery) mix inputs you cannot fully audit. None of this diagnoses sleep disorders. Loud snoring, witnessed apneas, or overwhelming daytime sleepiness belong in a clinic, not a blog comment section.
Training Decision Rule
Change training from multi-night duration trends + daytime function (RHR, mood, easy-run RPE). Do not change training from a single stage pie chart or a one-off score dip when total sleep and legs are fine. For race week, prioritize time in bed over optimizing “deep sleep” percentages.