Whoop Targets Improved Sleep Stage Detection

Wearables are good at telling when you are asleep and much worse at telling which stage of sleep you are in, and every sleep score and recovery score built on those stages compounds the inaccuracy.

Why sleep stages are so hard to get right

Sleep stages are defined by brain activity. A sleep lab measures them with polysomnography: electrodes on the scalp, eyes and chin, plus breathing and oxygen sensors, with a trained scorer labelling every 30 seconds of the night as wake, light (N1 or N2), deep (N3) or REM sleep. Human scorers make mistakes too, and even trained scorers agree on the stage only about 83% of the time.

A wrist wearable sees none of that. It has to infer brain states from heart rate, heart rate variability, movement and skin temperature, which change only loosely with sleep stage. Measure those slightly incorrectly, or use a slightly skewed algorithm, and one brand can very easily produce different results from another, even more so when the devices are worn on separate wrists.

Getting sleep stages right matters for the credibility of wearables because time in each stage is used to infer recovery, and so feeds sleep scores and sleep coaching. Stage figures also decide whether you trust the device at all. If the deep sleep number looks wrong to you, it is harder to believe the recommendations that flow from it.

Why each stage matters for recovery and wellbeing

A night runs through four or five cycles of non-REM and REM sleep. Deep sleep is concentrated in the first half of the night, and REM periods get longer toward morning, so a short night costs you mostly REM, and a broken first half costs you mostly deep sleep.

  • Deep sleep (N3) is the stage most closely tied to physical recovery and to waking up refreshed. Most of the night’s growth hormone, which supports tissue repair after training, is released in the first deep sleep periods.
  • REM sleep is tied to mental recovery: processing emotions, steadying mood and consolidating what you learned the day before. Short REM is linked to poorer emotional control the next day.
  • Light sleep (N2) fills about half the night and is not wasted time. Its bursts of brain activity, called sleep spindles, are linked to learning motor skills, which matters for technique in swimming, cycling and running.
  • Memory needs both: consolidation most likely requires both non-REM and REM sleep.

The uncomfortable truth for athletes is that total sleep is the biggest lever, stage proportions shift from night to night and with age, and nobody can train a single stage directly.

How Whoop proposes to fix it

Whoop’s idea is to add breathing to the signals it already uses.

  • Breathing in shortens the gap between beats and weakens each pulse; breathing out does the opposite.
  • The system measures four such changes on every beat: the gap to the next beat, the strength of the pulse, how steeply it rises, and how far it drifts from the average. From each, it rebuilds a breathing waveform and marks the start of every breath.
  • A small neural network decides which of those waveforms to trust at each moment, and the wristband’s motion sensor adds a second breathing signal when you are lying still.
  • From the breath timings, it scores respiratory rate variability, calculated much like HRV with the breathing equivalents of SDNN and rMSSD.
  • Low breathing variability points to deep sleep and higher variability to light sleep.
  • Deep sleep, which existing methods identify from heart rate, skin temperature, and motion, gives the cleanest readings, so the new approach uses it to calibrate the model. A trained classifier then scores REM and every non-REM stage.
  • The same score can also feed other compound recovery metrics.

The physiology behind it is sound. Breathing is slow and regular in deep sleep and irregular in REM, so a direct measure of breathing variability is a useful extra clue to the sleep stage.

The evidence in the filing

The application (US 2026/0283479 A1) was filed on 19 March 2025, published on 24 September 2026, and classified under detecting sleep stages and under wrist-worn devices. The sole inventor, Rahul Kumar Krishnaiah Sevakula, is a staff machine learning and AI scientist at Whoop, and employers normally own what their staff invent at work. The filing gives no accuracy figures and no comparison with polysomnography.

My take: promising, but it needs proof.

I think this shows promise. Breathing variability is a genuinely different signal from heart rate; it comes free from the sensor Whoop already has, and on paper it helps where wearables struggle most: telling deep sleep from light sleep. But it needs proper scientific validation against polysomnography, with stage-by-stage results published, or it risks becoming yet another poor way to detect sleep stages that looks great on a chart.

That said, there is a hard ceiling. Classifying sleep stages is inherently difficult when even the gold standard is not that accurate. In the American Academy of Sleep Medicine’s inter-scorer study, more than 2,500 trained scorers agreed with the reference only 82.6% of the time overall, 67.4% for deep sleep and 63.0% for N1. A wearable cannot beat the reference it is tested against, so the fair question for Whoop is how close it gets to human scorers, not whether it is perfect.

Quick answers

Can Whoop detect sleep stages from breathing?
Whoop has filed a patent application for a method that reads breathing from its heart rate sensor and uses breathing variability to estimate sleep stage. It has not said the method is in its app, and the filing gives no accuracy results.

How accurate is wearable sleep stage tracking?
Less accurate than the sleep lab, which itself is imperfect: trained human scorers agree on sleep stage about 83% of the time, and only about two-thirds of the time for deep sleep. Wearables are usually better at total sleep time than at individual stages.


Related: smartwatches could learn how well you feel you slept, Whoop’s patent for GPS-free VO2 max, the Whoop 5.0 and MG review and the Whoop hub.

Last Updated on 18 November 2026 by tfk the5krunner


Small things I always use

Health, recovery and sleep

  • Kineon MOVE+: Red and infrared light for sore knees, Achilles and other joints after hard training. I use it.
  • FreeStyle Libre 3 Plus: Shows how my blood sugar responds to diet. Occasional use.
  • hDrop: Measures my sweat loss in summer and indoors. I use it.
  • Fourth Frontier Zone: Checks my heart under load in hard sessions.
  • Oura Ring 5: My partner’s recovery tracker of choice.
  • Eight Sleep: A cooling mattress cover my partner and I use nightly.

Watches and sensors

On the bike

Home and indoor training

  • NPE Runn: Makes my basic treadmill smart, sending more accurate speed and distance to my watch and Zwift. I use it.
  • Google Wifi: Keeps Zwift connected in my pain cave. I use it.
  • Apple TV 4K: Zwift on the big screen. I use two.
  • Google Nest Doorbell (wired, 2nd gen): My video doorbell, running in both Google Home and Apple Home via Starling. I use it.


Reader-Powered Content

Buy me a coffee

This content is not sponsored. It’s mostly me behind the labour of love, which is this site, and I appreciate everyone who supports it.

Support the site: Follow (free, fewer ads) · Subscribe (paid, ad-free) · Buy Me A Coffee

Written by real people, with AI assistance. Fact-checked and verified for originality. See the Editorial Policy. FTC: Affiliate Disclosure — some links pay commission. As an Amazon Associate, I earn from qualifying purchases.

Leave a Reply

Your email address will not be published. Required fields are marked *