TraviaTechPie Review

The Story

Here’s a thing almost nobody tells you when you strap a watch on before bed: your smartwatch never actually watches you sleep. It guesses.

That’s not an insult to the technology. It’s just how the physics work. A hospital sleep study — the real gold standard, called “polysomnography” or PSG — sticks electrodes on your scalp and reads your brain waves directly. That’s the only way to know, for certain, which sleep stage you’re in. Your watch has none of that. It’s sitting on your wrist, inches away from your brain, trying to reverse-engineer what’s happening upstairs from three clues: how much your arm moves (an accelerometer), how your heart beats (an optical sensor shining green light through your skin, called “PPG”), and on some models, blood oxygen and skin temperature.

So when the app in the morning shows you a tidy bar chart — 1h 12m of deep sleep, 1h 40m of REM, a crisp little hypnogram — understand what you’re looking at. That chart is a model’s best inference from wrist motion and heart rhythm. It is not a measurement of your brain. Keep that one fact in your head and everything else about sleep tracking suddenly makes sense.

So what is it actually good at? More than you’d think, in the places that matter most day to day. The single most validated thing wearables do is tell sleep from wake. Across device-versus-PSG studies, sensitivity for detecting that you’re asleep runs roughly 91–96% across studies. That means the top-line numbers — roughly when you fell asleep, roughly when you woke, and your total sleep time — are usually in a believable ballpark. If your watch says you got five and a half hours, you probably didn’t get eight. That’s genuinely useful, and it’s the number most people should care about anyway.

Where it gets shaky is the pretty part: the stages. Splitting the night into light, deep, and REM is exactly the job that really needs brain waves, and the wrist just doesn’t have them. In validation research, agreement between wearables and PSG on staging lands in the “fair to moderate” range — statisticians use a measure called Cohen’s kappa, and consumer devices tend to score somewhere around 0.2 to 0.6 depending on the device, where 1.0 would be perfect. Deep sleep is a particular trouble spot: some devices undercount it by tens of minutes on a given night; others overcount by similar or larger margins — direction depends on brand/algorithm. And stages get shuffled: a chunk of deep or REM quietly labeled as “light.” Newer models with heart-rate-variability input tend to score better on staging, but the improvement varies from device to device, and no consumer device has matched the lab. Treat your stage breakdown as a rough sketch, not a transcript.

There’s a quieter set of numbers that’s more trustworthy than the stage chart, and most people ignore it: your resting heart rate and your “HRV” (heart rate variability — the tiny beat-to-beat timing changes that reflect how your nervous system is doing). Those come straight off the optical sensor as trends, and trends are what wrists are good at. A resting heart rate that creeps up for three nights, or an HRV that dips, may appear before you consciously feel run-down. That signal is real even when the deep-sleep bar is fiction.

Now, how do you make the whole thing more accurate? Most of the error people blame on the device is actually a fit-and-setup problem. A few habits do most of the work:

• Snug the band, correctly. The optical sensor needs skin contact with no daylight gap. Wear it a finger’s width above the wrist bone, tight enough that it doesn’t slide when you rotate your arm, loose enough that it isn’t cutting in. Too loose is the number-one cause of garbage heart-rate data at night.
• Actually turn on sleep mode. Set a sleep schedule or bedtime in the app so the device knows the window to watch. Guessing the window from scratch every night is harder and less accurate.
• Have it on before you’re drowsy. Put the watch on at least 20 to 30 minutes before bed. Sleep-onset detection leans on the transition from awake-and-still to asleep, and it reads that better with a running start.
• Don’t trust naps. Most wrist trackers are tuned for a full night. A 25-minute nap may not register as sleep at all, or gets logged oddly. Some devices have a dedicated nap mode — if yours doesn’t, don’t read too much into daytime numbers.
• Check the sensitivity setting. Several platforms let you dial sleep detection more or less aggressive. If your watch keeps insisting you were awake during quiet reading, or misses real wake-ups, that toggle is worth a look.
• Keep the firmware current. Sleep algorithms get retrained and shipped in software updates far more often than people realize. An old build can be measurably worse than the same hardware six months later.

One more decision hides inside all of this: what you wear it on. If sleep is your main reason for tracking, a bulky watch is fighting you. It’s heavier on the wrist, and — the real killer — it usually needs charging daily, which is to say it needs charging exactly when you’d otherwise be wearing it to bed. A lightweight band or a smart ring sidesteps both problems: barely-there on the wrist or finger, and multi-day battery, so it’s actually on your body every night. That consistency matters more than a spec sheet, because a tracker you took off to charge measured nothing. (I got into this trade-off in the smartwatch vs. fitness tracker breakdown, and it applies double at night.)

The Takeaway

The mistake I see most is people reading their sleep app like a report card — obsessing over a single night’s deep-sleep bar as if it were a lab value. It isn’t. The stage breakdown is the least reliable thing on the screen, and one night of anything is noise.

Read it the other way around. Look at trends across a week, not verdicts on a night. Total sleep time and consistency of your bed-and-wake times are the numbers that earn their keep, because those are what the device measures well. Resting heart rate and HRV moving over several days will tell you more about how recovered you are than any hypnogram. And that “sleep score” a lot of watches hand you? It’s a blended, brand-specific composite — useful against your own baseline, close to meaningless compared to your friend’s. Watch your own line move; don’t compete on the absolute number.

This connects to something I keep coming back to across these wearable pieces — in the health sensors explainer and the ECG-and-heart-rate accuracy piece too. The wrist is a fantastic trend machine and a mediocre instrument. It’s brilliant at “is this different from your normal?” and weak at “here is the precise clinical value.” Sleep tracking is that same story in the dark.

And the hard line: a smartwatch cannot diagnose a sleep disorder. If you snore heavily, gasp awake, feel wrecked after a full night, or your numbers stay ugly for weeks, that’s a doctor’s job, not an app’s. The watch’s real value is spotting a change worth mentioning — then handing the question to someone with electrodes.

This article is for informational purposes only and is not medical advice.

TL;DR: Your watch guesses sleep from wrist motion and heart rate, not brain waves. It nails asleep-vs-awake and total sleep time (roughly 91–96% sensitivity for sleep/wake across studies), but stage splits (light/deep/REM) are only fair-to-moderate vs. a lab — treat them as a sketch. Fit the band snug, set sleep mode, wear it before bed, update firmware. Read weekly trends, not single nights; lean on resting HR and HRV. A light band or ring beats a watch here because battery lets you wear it every night. It can’t diagnose anything — persistent problems go to a doctor.

Photo: Indra Projects / Unsplash