We Looked at 29,388 Nights: Your Deep-Sleep Number Isn't Your Recovery Score
Every wearable tells you to chase deep sleep. We compared 29,388 nights to each person's own normal night. More deep sleep didn't mean a better morning recovery reading. More REM and not waking up did.
"Chase Your Deep Sleep" Is the Wrong Advice
Open any sleep tracker and deep sleep gets the spotlight. The bold number, the colour, the little "restorative" tag. The story everyone repeats is that deep sleep is when your body repairs itself, so more deep sleep means a better day tomorrow.
I never fully bought it. The athletes I work with have wildly different deep-sleep numbers and recover at rates that don't match them at all. So I went and looked at the data we actually have.
First, what we are measuring. The "recovery reading" I keep mentioning is HRV, short for heart rate variability. It is the overnight number your watch turns into a recovery or readiness score. When it is higher than your usual, your body is rested and in a good state. When it is lower, you are more run down. It is the closest thing a wearable has to a daily recovery gauge.
athletedata has 29,388 nights where we can see how the night broke down (how much was deep, REM, light, and time spent awake) sitting right next to that morning HRV reading, from 392 people on Oura, WHOOP, Garmin, Apple Watch and the rest. That is enough to check the advice properly instead of repeating it.
The short version: more deep sleep did not come with a better recovery reading. More REM and not waking up did. And the popular "deep sleep is the recovery stage" picture turns out to be a trick of comparing different people to each other.
How We Looked At It (And Why It Matters)
Here is the trap in almost every sleep stat you read. Deep sleep drops steeply as you get older, from about a fifth of the night in your twenties to a sliver by middle age, and how much you get is partly down to your genes. So if you throw everyone into one pile, the people with lots of deep sleep are mostly the younger, fitter ones, who also have a higher recovery reading for reasons that have nothing to do with last night.
Compare strangers to each other and you will "prove" that deep sleep drives recovery, when all you have really shown is that being young helps both.
To get around that, I compared every night to that same person's own usual night. "More deep sleep" means more than they normally get. "A better reading" means a better morning than they normally get. Then I lined the two up, and I did it for resting heart rate, readiness, and recovery score too, so I was not leaning on a single number.
This is the same compare-you-to-yourself method the serious research uses, and it is the only way to tell "this night moved my recovery" apart from "this is just how I sleep."
Finding 1: Deep Sleep Was the Flat Line
On the nights someone got more deep sleep than their own usual, their morning recovery reading was no better. The line barely moves, and if anything it drifts slightly down. Resting heart rate, readiness, and recovery score all said the same thing. Compared to your own normal, the deep-sleep number simply did not track how recovered you were.
That surprised me, so I ran the lump-everyone-together version as a check. Across people, deep sleep does look like the winner, exactly the relationship the apps lean on. The moment I compared each person to their own usual, it vanished. The crowd average and your own nights point in opposite directions, and only one of them is about your training.
Finding 2: REM and Staying Asleep Were the Lines That Moved
The two things that did line up with a better reading were getting more REM than usual, and spending less of the night awake.
A night with more REM came with a higher recovery reading, a lower resting heart rate, and a better readiness score. A night with more time awake came with the opposite. Sleeping longer than usual helped too, though less than most people assume.
If you have watched your resting heart rate creep up during a heavy block, a restless, broken night is the same thing on a one-night timescale.
What This Does Not Mean
Here is where I have to be careful, because the obvious read is wrong.
Minute to minute, your body actually does its deepest recovery during deep sleep, not REM. So "more REM, better reading" is not REM doing the repair work. Getting plenty of REM is a sign that you had a calm, unbroken night that ran all the way through its cycles. REM is the first thing that gets cut when something is off, a late hard session, stress, or a couple of drinks before bed. So treat REM and a night without much waking as signs of a good night, not buttons you can press.
Two more things that change how you should use this. It is a same-night reading. When I checked whether last night's breakdown predicted the next morning's reading, there was nothing there. And deep sleep is the single hardest stage for a watch or ring to measure. Studies that compare them to proper lab sleep tests find they mislabel a quarter to a half of deep sleep. So part of why deep looked flat is that it is the shakiest number on your screen to begin with.
I am not saying deep sleep does not matter for your body. I am saying the deep number on your watch does not track your recovery, and it is unreliable on top of that.
This Lines Up With the Research
None of this argues with the science, it agrees with it. A 2025 study found the recovery signal showed up in REM, not in deep sleep (Morehouse et al.). Broken or short sleep reliably pushes your heart rate up and your HRV down (Schlagintweit et al.). And deep sleep is mostly set by your age and genes, so it has little room to swing from night to night.
The research was never the problem. The problem is what happens when careful science gets squashed into one gauge. "Deep sleep helps your body repair" becomes "your deep-sleep score is your recovery," and the part that actually helps you gets lost. The same flattening happens with WHOOP recovery and Oura readiness scores, which bundle a lot of moving parts into one daily figure.
What I Would Actually Do With This
Stop grading your night by the deep-sleep bar. The thing that lines up with your recovery, and the thing your watch measures most reliably, is how much of your time in bed you actually spent asleep. Watch that instead.
If your REM looks low for a stretch, treat it as a nudge to look at what is wearing you down, late training, alcohol, travel, rather than a number to chase. And if your deep number is low on a night you slept straight through and woke up feeling good, I would not give it a second thought.
The honest version of this question is not what 392 strangers average out to. It is what your own few hundred nights say. That is the chart sitting in your own data, and it is the one worth reading. If you want it built from your nights instead of a crowd average, that is the kind of thing athletedata does with your training and sleep data.
athletedata connects your Garmin, Oura, WHOOP, Apple Watch and Strava, compares every night to your own normal, and coaches you over WhatsApp or Telegram. It reads your own data instead of the crowd average. See how it works.
