Sleep and Wearable Data: What Your Watch Measures

Your wearable reports a sleep score, but that number comes from estimated stages, not measured ones. Polysomnography, the lab gold-standard for sleep research, uses brain activity, eye movement, and muscle tone to know what stage you're in. Your watch uses heart rate patterns and accelerometer data to guess. The gap between guessing and knowing matters when you're trying to read what last night delivered for your recovery. This guide walks through what wearables measure, which metrics predict performance, and how to read the data without overthinking the noise.

What consumer wearables measure: the honest limits

Every modern wearable claims to detect sleep stages: light, deep, REM, awake. None of them do it well. A validation study comparing six commercial wearables to polysomnography found Cohen's kappa values ranging from 0.21 to 0.53, meaning agreement ranged from fair to moderate. Apple Watch Series 8 performed best (κ = 0.53). Garmin Vivosmart ranked lowest (κ = 0.21). For context, kappa above 0.6 is generally read as substantial agreement. No consumer device reached that threshold.

What wearables do well: all devices detected more than 90% of sleep epochs, meaning they can tell if you're asleep or awake with solid accuracy. The problem is staging. Sensitivity (catching sleep) was consistently high at 90-96%, but specificity (correctly catching the minutes you were awake) was only 29-52%, so devices often score wake time as sleep. Your watch is good at knowing you slept. It's poor at knowing what kind of sleep, and it flatters your total.

Sleep duration is different. When tested against lab measurement, wearables showed far better accuracy for total sleep time than for staging, with sleep efficiency estimates varying widely but generally useful for trends. The lesson: trust the hour counts, not the stage percentages on any single night.

What sleep stages are

Sleep cycles through two main types: non-REM (light and deep) and REM (rapid eye movement). Each has a different job.

Light sleep (N1 and N2) is when your body transitions from wake into deeper recovery. Heart rate begins to drop. This stage makes up about 50-60% of total sleep and is harder to detect via wearable (it looks similar to wakefulness on heart rate alone).

Deep sleep (N3, slow-wave sleep) is the stage doing the physical repair work (why each stage matters for adaptation is covered in the sleep guide). Detection-wise it is the hard one: deep sleep is rare (typically only 10-20% of the night) and wearables struggle to isolate it from light sleep because optical wrist sensors cannot measure brain waves.

REM sleep is when your brain consolidates skills and memories. Heart rate elevates and variability increases, which makes it harder to distinguish from light sleep for algorithms looking at cardiac signals. Most staging algorithms show elevated heart rate variability in both REM and light sleep, an overlap that limits how reliably a wearable can tell the two apart from heart rate data alone.

The wearable challenge: it can estimate these stages by looking at patterns in heart rate, movement, and breathing, but the patterns overlap. A light-sleep heart rate looks like an awake heart rate. A REM heart rate looks like a light-sleep heart rate. This is why you get stage percentages on your watch that shift wildly from night to night despite no real change in how rested you feel.

Which metrics matter for performance

Duration beats staging as a coaching signal. Total sleep time is the more consistently reported predictor of recovery and performance in the sleep-and-athletics literature, while stage-distribution effects are less established and still being studied. A systematic review of objective sleep measures in athlete populations reports duration and architecture findings separately without ranking one over the other for predicting recovery, but in practice an athlete sleeping 7 hours of fragmented light sleep will usually outperform one sleeping 5 hours of optimal-stage distribution. Wearables can measure duration. They guess stages.

Regularity matters, too; why consistency beats weekend catch-up is covered in the sleep guide. The wearable angle: your device tracks timing automatically if you check the timestamps. They show this as a column chart in the companion app (Garmin Connect, Health app) but don't always highlight it as the signal it is.

Heart rate and HRV during sleep reveal autonomic state. Resting heart rate during sleep and nocturnal HRV generally settle as parasympathetic tone (the "rest" branch of your nervous system) rebuilds through the night. Chronically elevated nighttime RHR or suppressed nocturnal HRV (the kind that persists across multiple nights) signals that recovery is incomplete. This is more useful than any stage percentages. Both Garmin and Apple Health expose these values in the data stream and the coach reads them. A single bad night of elevated RHR is noise. Five straight nights trending up is signal.

How Garmin and Apple Health measure sleep

Garmin reads heart rate, heart rate variability, motion (via accelerometer), and respiration to estimate sleep stages. The algorithm outputs a "Sleep Score" (0-100) and breaks down stage percentages. Garmin's own documentation acknowledges the limitation: the staging is consumer-grade estimation, not clinical-grade measurement. The platform is solid for trend tracking because the same algorithm applies consistently across your data.

Apple Health collects sleep duration and estimated stages from the iPhone and Apple Watch sensors, primarily using accelerometer data (motion) plus heart rate when available. Both platforms rely on wrist optical heart rate plus motion, and in head-to-head validation neither reaches clinical staging accuracy; treat both as duration-and-trend tools. On the iOS app, Apple Health sleep data syncs with the Movement Rebels data stream, and the coach reads duration + HRV trend from the same window.

Neither device measures deep sleep accurately. Neither can distinguish REM from light sleep reliably. Both are useful for duration and for seeing when your HRV or resting heart rate deviates from baseline. Use them that way.

Accuracy and what it means for your training

A large multicenter study testing 11 sleep trackers found that sleep stage classification showed F1 scores ranging from 0.26 to 0.69, with top devices achieving only moderate agreement. This means even the best consumer device is wrong about 30-40% of the time on staging. Duration estimates are generally more reliable than stage estimates, though exact accuracy varies by device and by night.

The practical implication: your watch is reliable at telling you that you slept 7 hours last night. It is not reliable at telling you that 20% was deep sleep. If you see your deep-sleep percentage drop from 18% to 8%, do not assume your sleep got worse on that specific night. That swing could be noise from the staging algorithm. But if deep sleep averages 8% for a month, then 5% for the next month, that trend is meaningful.

The right frame: treat single nights as data points, not diagnoses. Trends across 7-14 days are where the signal lives. This is also why the morning brief surfaces rolling averages of HRV and RHR rather than single-night values. The coach is trained to ignore the noise and watch the slope.

Wearables for sleep debt and recovery clusters

A pattern worth watching in your own data: declining sleep score, rising resting heart rate, and dropping HRV across the same week can point to under-recovery before performance drops. This is not a single validated marker that behaves the same for everyone. A study following recreational runners through a two-week training overload block found meaningful differences between individuals in how sleep and nightly recovery metrics tracked training stress, with subjective recovery ratings shifting more consistently than the objective sleep and HRV data at the group level. Track your own trend rather than assuming this cluster reads the same way for you as it does for someone else.

When you see the pattern (sleep score declining over three nights even though total hours stay the same, RHR elevated above your baseline, HRV trending down (see the HRV baseline guide for sizing that drop against your own 28-day baseline)), it's worth treating as a flag that recovery may be incomplete. The cause could be training stress, life stress, late caffeine, insufficient nutrition, or a combination. The solution is usually a deload or a shift in session intensity, not more sleep alone (though more sleep helps). The coach adjusts the session accordingly.

Garmin surfaces related signals through Body Battery and HRV Status. Apple Health has no native recovery index, but the Movement Rebels coach reads all three signals (sleep, RHR, HRV) from Apple Health and flags the pattern in the morning brief.

How to use your sleep data

Read the trends, not the numbers. Set a rolling average. Most apps show a 7-day view; use that instead of chasing individual nights. When your sleep score drops, check two things before assuming it's bad:

1. Did the duration change? If you slept the same hours but the score dropped, the staging algorithm shifted. That's noise. Move on.
2. Is RHR or HRV also off? If sleep score drops and your resting heart rate is elevated or HRV suppressed the next morning, something real happened. Training stress, alcohol, caffeine, insufficient fuel, or poor temperature all contribute. If sleep score dropped but RHR and HRV are normal, the scoring just fluctuated.

Track timing. Consistency matters more than total hours. If your 6.5-hour sleep at the same time every night shows high HRV and low RHR, that's better recovery than 8 hours at drifting times. The wearable shows this in the timestamp history (under "Sleep History" in Garmin Connect or the Health app). Check it weekly; the coach reads timing consistency along with the rest of the trend.

Use the coach data. If you're connected to Movement Rebels and Garmin or Apple Health, the coach reads all of this (duration, timing, HRV, RHR) and adjusts your plan based on the cluster, not a single metric. You do not need to interpret the staging percentages yourself. The coach does it for you.

How Movement Rebels fits

The coach reads your Garmin or Apple Health sleep data daily: total sleep time, timing, resting heart rate, heart rate variability, and estimated stages. It does not use the stage percentages to diagnose your sleep. It uses duration + timing + HRV + RHR trend to assess recovery state.

When sleep score drops or HRV suppresses despite adequate hours, the coach reads context: what you trained yesterday, whether you logged a late hard session, what your nutrition looked like, stress signals from your HRV trend, and what is scheduled today. The adjustment playbook itself lives in the sleep guide; this guide's point is that the adjustment is driven by the data trend, not the nightly stage pie chart.

The coach also integrates sleep recovery with your training load and readiness metrics, so the plan adapts to both performance markers and recovery capacity. That integration across domains (sleep plus training plus HRV plus nutrition) is where wearable data becomes actionable rather than decorative.

Start with one habit: check your rolling 7-day sleep average and timing consistency once a week, and notice when RHR or HRV deviate from baseline on the same morning your sleep score drops.