For most of my adult life I have known exactly how well I slept within about four seconds of waking up. No instrumentation required. I would open one eye, assess my willingness to speak to another human being, and issue a verdict. Good sleep: cheerful, borderline insufferable. Bad sleep: monosyllabic, dangerous to approach before coffee. This system had a sample size of roughly fifteen thousand nights and, as far as I could tell, a fault rate of zero.
Then I strapped a small black rectangle to my wrist and let it start marking my homework.
For a while this was thrilling in the way all new data is thrilling. Heart rate, HRV, something called a Sleep Score, all beautifully rendered on an app with the calm, confident typography of a brand that has never once been wrong about anything. I woke up each morning and checked my phone before I checked my own body, which is a sentence I did not enjoy writing.
Then the disagreements started.
I would wake up feeling like I’d been poured from a jug, ready to take on the day and possibly a small argument, and the app would inform me, with the quiet authority of a disappointed schoolteacher, that I had scored 61 and should consider an early night. Other mornings I would surface feeling like a man who had personally lost a fight with a taxi, and the app would cheerfully award me an 89 and tell me I was primed to perform.
My first response was denial. The band was too loose. The band was too tight. I’d slept on my charging arm. I tried every excuse a man can try before admitting that a piece of silicone might understand his own nervous system better than he does.
My second response, which I am less proud of, was worse. I started believing it. If the score said poor, I would spend the day quietly agreeing with it, manufacturing tiredness I hadn’t actually felt ten minutes earlier, as though the number were a diagnosis rather than a guess. I had, without noticing, handed over the one instrument I’ve always trusted — my own sense of how I feel — to a company whose real product is a monthly subscription.
This wasn’t even new ground for me, which makes it worse. I’ve written before about experimenting with HRV to catch overreaching before it wrecked a block, and about the slightly alarming research showing that elite athletes — professionals, whose entire job is recovering properly — average a couple of hours less sleep than they themselves say they need. Both times the lesson was the same: the boring, unglamorous number is the one worth watching. I still somehow needed to relearn it, this time from a Sleep Score with main character energy.
So I went looking for what these devices are actually measuring, and how well, because it turns out that’s a question with an answer, not just a marketing page.
Here is the flat version, no jokes attached. Consumer wearables are genuinely good at two things: total sleep time and overnight heart metrics. Independent validation studies comparing Oura, WHOOP, Garmin and Apple Watch against polysomnography — the full lab test with electrodes on your skull — show resting heart rate and HRV agreeing with clinical-grade measurement extremely closely, often above 0.97 on the statistical scale researchers use for “basically the same number.” Total sleep time is similarly solid across most devices.
Where it falls apart is sleep staging — deciding how much of the night was deep, light or REM. That’s the part the Sleep Score is quietly built on, and it’s the part these devices are worst at. Accuracy for correctly identifying which stage you’re in typically sits around 65 to 70 per cent, because a watch has no idea what your brain is doing. It only has your movement and your pulse, and it’s making an educated guess about the rest. Polysomnography reads actual brainwaves. Your wrist reads fidgeting.
Which means the headline number I’d been trusting most — the Score — is built substantially from the measurement these devices are least equipped to take, while the numbers they’re genuinely excellent at, resting heart rate and HRV trend, sit quietly further down the app, undramatized, un-gamified, and largely ignored by me because they don’t come with a colour and a verdict.
I’d been furious at my body for contradicting a number generated, in significant part, by a guess.
This is, I’ll admit, a very familiar shape of idiot for me to be. I spent years assuming a Garmin file was proof of fitness rather than a record of effort. Now I’d found the sleep equivalent: assuming a single blended score was proof of recovery, when the useful information was two unglamorous trend lines sitting underneath it, doing their job without any interest in being noticed. It’s basically the same mistake Dan Plews has spent his career trying to talk athletes out of. He’s the physiologist who helped New Zealand’s Olympic gold-medal rowers make sense of exactly this kind of data, and his line has always been that a single day’s reading is noise — the trend over weeks is the only part that tells you anything true.
I haven’t thrown the device away. That would be the sort of dramatic gesture that feels good for a day and achieves nothing. What I’ve actually done is downgrade the Score to entertainment — a horoscope with a battery — and started paying attention to the resting heart rate and HRV trend over weeks, not the nightly verdict, alongside the frankly ancient technology of asking myself how I feel. Somewhere between the fifteen-thousand-night gut instinct and the four-figure gadget on my wrist is an actual answer, and it turns out it was never going to arrive as a single number with a smiley face attached to it.
I still check the app first thing every morning. I just no longer let it tell me how my day is going to go before I’ve had the chance to find out for myself.
This is exactly why recovery isn’t a footnote in my training plans — it’s built in, structured around the trends that actually mean something rather than the nightly verdict that doesn’t. I’m aware that sentence is the bit where I try to sell you something. That’s because it is one.