How Accurate Are the Sleep Stages From Your Watch or Ring?

Your watch says you had 48 minutes of deep sleep last night and 1 hour 52 minutes of REM. It’s tempting to chase those numbers. Before you do, it’s worth knowing how they’re made, and how close they come to what a sleep lab would measure.

This article is general information, not medical advice. If you think you have a sleep problem, such as loud snoring with pauses in breathing or long-lasting insomnia, talk to a health professional.

How Sleep Stages Are Really Measured

The reference method is polysomnography (PSG), done in a sleep lab or sometimes at home. Sensors on your scalp record brain waves (EEG), with others for eye movements, muscle tone, breathing, and heart rhythm. A trained scorer then labels each 30-second stretch of the night as one of five stages:

• Wake
• N1, light, transitional sleep
• N2, which makes up most of non-REM sleep
• N3, deep or slow-wave sleep
• REM, rapid eye movement sleep, when most vivid dreaming happens

Stages are defined by what the brain is doing. That’s the key point: a wearable can’t see your brain.

How a Wearable Estimates Them

A watch or ring measures what it can reach from your wrist or finger: movement, heart rate, heart rate variability, and sometimes skin temperature and breathing. An algorithm, trained on nights recorded with PSG, then estimates which stage each stretch of your night most likely was.

Apple, for example, describes its sleep stages as coming from a classifier built mainly on the watch’s accelerometer, which picks up body movement including the small motions of breathing. It was trained and tested against PSG recordings. Apple groups the stages into Core, Deep, and REM, plus Awake. Other brands use their own names and their own methods.

What the Studies Found

Wearables are good at one thing and weaker at another.

Sleep versus wake: good at sleep, weak at wake. Chinoy and colleagues tested seven consumer sleep trackers against PSG in a sleep lab, over three nights with 34 healthy young adults. The devices recognised sleep very well (a sensitivity of 0.93 or more), but were much weaker at recognising wake (a specificity of 0.18 to 0.54). Lying still awake often looks like sleep to a device. Most did worse on nights when sleep was disrupted.

Stages: mixed. In the same study, every device that reported stages differed from PSG on light sleep, mostly overestimating it. Some overestimated deep sleep and some underestimated REM. Most devices missed 30% to 50% of deep and REM sleep when compared stretch by stretch.

Devices change with every generation, and newer ones may do better than those tested. But the basic limit remains: they estimate brain states from the body.

What’s Worth Tracking

Some sleep numbers are more trustworthy than others. Roughly, from most to least reliable:

1. When you went to bed and got up. Timing is the easiest thing for a device to get right, and it shows at a glance how regular your nights are.
2. Time asleep, over weeks. Night-to-night numbers wobble, but a weekly average from the same device is a fair trend. The American Academy of Sleep Medicine and the Sleep Research Society recommend that adults sleep 7 or more hours a night on a regular basis.
3. Time awake in the night. Useful as a trend, but expect it to be underestimated, since quiet wakefulness looks like sleep.
4. Deep, light, and REM minutes. Treat them as rough estimates. A change in your device’s deep sleep from one night to the next may be the algorithm, not you.

And how you feel matters too. A rating of your energy each morning is a simple check on what the numbers say.

Tracking Sleep in Kaizora

Kaizora’s Sleep page holds each night or nap, with its stages when your device gives them.

• Import Apple Health. Each night comes in with its stages: Apple’s Core sleep is kept as light sleep, and Deep, REM, and Awake keep their names. See Importing From Apple Health.
• Or log a night yourself with when you went to bed and got up. Stages come only from imports, since no one can time their own.
• Every number is its own metric: Time in Bed, Time Asleep, Time Awake in Bed, Light Sleep Time, Deep Sleep Time, and REM Sleep Time. A night counts toward the day you wake up.
• No double counting. If two sources both recorded the same night, a day’s total takes one of them, not both added together. See Why Your Devices Disagree.
• Set a habit of at least 7 hours asleep a night and watch your streak. See Habits.
• Compare with how you feel. Log your Energy Level and Mood on the Wellbeing page, each on a scale of 1 to 5, and chart them against Time Asleep on Correlations. See Comparing Metrics.

The Short Version

• Sleep stages are defined by brain activity, which only PSG measures.
• Wearables estimate stages from movement and heart signals. They’re good at detecting sleep, weaker at detecting wake, and mixed on stages.
• Trust timing and weekly time asleep the most, and stage minutes the least.
• Use one device, follow weekly averages, and check the numbers against how you feel.

If you’d like your nights next to your mood, energy, and everything else, you can try Kaizora for free.

Sources

• Chinoy ED et al., “Performance of seven consumer sleep-tracking devices compared with polysomnography,” Sleep 44, no. 5 (2021): zsaa291. doi:10.1093/sleep/zsaa291
• Watson NF et al., “Recommended amount of sleep for a healthy adult: a joint consensus statement of the American Academy of Sleep Medicine and Sleep Research Society,” Sleep 38, no. 6 (2015): 843–844. doi:10.5665/sleep.4716
• Apple, Estimating Sleep Stages from Apple Watch (2023, updated October 2025).