The second chapter of this summer story ended with a spark — a tiny, improbable spark — ignited by a YouTube video that should have drowned in the ocean of algorithmic garbage. But it didn’t. It cut through the noise like a flare in a storm, and before I even realised what was happening, I had clicked. And that click, that innocent, almost accidental gesture, became the doorway into everything that followed.
And today’s post begins exactly where my story had paused: with this video. Because this wasn’t just any video. This wasn’t about calories, or steps, or “getting moving.” It was about sleep — that old obsession, that unfinished business from 2015. But more importantly, this video wasn’t the usual marketing fluff. It was science. Actual evidence. Validation papers. Hundreds of studies. Real data. Not influencer nonsense. Not the “Top 10 gadgets you NEED in 2026” plague that has taken over YouTube. This was different. This was serious. This was the first time in years that someone, somewhere, seemed to be speaking my language.
What this guy — The Quantified Scientist — did was not a review. It wasn’t an unboxing. No. This was something else entirely. This was someone who had gone into the literature — the real literature — the electroencephalogram (EEG) kind, the kind written by people who argue about electrodes, sleep epochs, and signal noise. He pulled out every validation study he could find 1. And then he did something that made the scientist in me sit up straighter: he extracted the numbers. Sensitivity. Specificity. Stage‑tracking accuracy. Agreement metrics. All of it.
He dumped everything into a spreadsheet — a monstrous, glorious spreadsheet — and instead of drowning in it, he tamed it. He grouped devices by brand. Separated old algorithms from new ones. Split healthy sleepers from people with sleep disorders. Averaged results across studies so that one rogue paper wouldn’t hijack the truth. It was the kind of meticulous, borderline‑obsessive work I love. And once he had that mountain of data under control, he turned it into something beautiful: graphs. Not marketing graphs. Not glossy infographics. Real scientific plots — the kind that care only about truth.
The first graph was a sensitivity‑versus‑specificity scatter plot — the holy battlefield of sleep‑wake detection. On the X‑axis (the horizontal one at the bottom), Sensitivity: the ability to detect sleep. On the Y‑axis, Specificity: the ability to detect wakefulness. Each point represented a device, its data averaged across studies. A perfect device — none exist yet — would sit proudly at 100% for both; 100 % agreement with the scientific way to track sleep (i.e., with Polysomnography). But that’s not what the plot showed. Instead, the devices lined up along a diagonal, as if obeying some unwritten law of mediocrity: from relatively high Specificity (topping out around 80%, not 100%) but lower Sensitivity (around 90%), all the way down to abysmal Specificity (20%) paired with high Sensitivity (100%). Why such a trade‑off? I don’t know.
Polysomnography (PSG; which has nothing to do with a French soccer team) is the gold standard for measuring human sleep. It is the method against which all wearable and home‑based sleep technologies are validated. Unlike consumer devices that infer sleep stages from wrist movement and heart‑rate patterns, PSG directly measures the physiological signals that define sleep itself. These include:
EEG (Electroencephalography) — measures brain waves and is the primary determinant of sleep stages.
EOG (Electrooculography) — detects eye movements, essential for identifying Rapid Eye Movement (REM) sleep.
EMG (Electromyography) — measures muscle tone; REM sleep is marked by near‑complete muscle atonia.
ECG (Electrocardiography) — monitors heart rhythm.
Respiratory sensors — track airflow, breathing effort, and oxygen saturation.
Additional channels — may include snoring microphones, limb movement sensors, and body position monitors.
This process is labor‑intensive and requires specialized expertise, contributing to the high cost of PSG and limiting its availability to accredited sleep centers. But because PSG captures the actual physiological signatures of each sleep stage, it can distinguish N1 (light sleep) from N2 (light sleep with sleep spindles and K‑complexes), N2 from N3 (deep, slow‑wave sleep), and REM from all non‑REM stages with high precision — something wrist‑based wearables fundamentally cannot do.
Back in 2015, that was exactly my frustration — watching real sleep science get eclipsed by wrist‑worn wishful thinking.
Modern wearables rely on: Accelerometers (movement), photoplethysmography (PPG 2) sensors (Heart Rate and Heart Rate Variability), and Algorithms trained on (healthy) population averages. They attempt to infer sleep stages from these indirect signals. But sleep stages are defined by EEG patterns — something wearables do not measure. As a result, wearable sleep‑stage accuracy is only 50–65% compared to PSG 3. Deep sleep is often overestimated because stillness + low heart rate can occur in both N2 and N3. REM sleep is frequently misclassified because wearables cannot detect eye movements or muscle atonia. Quiet wakefulness is often mistaken for sleep, inflating total sleep time.
All the wearable can do is make educated guesses about your sleep stages, and those guesses are wrong about half the time (based on the wearable).CuriousCatalyst
Wearables may eventually approximate PSG for certain metrics, but they will not replace it until they can reliably measure brain activity, eye movements, and muscle tone — the core physiological signatures of sleep.
By 2026, the landscape had shifted — but not nearly as far as the hype suggested.
For a deeper dive into the evidence behind wearable sleep‑stage accuracy, this article offers a clear, research‑based overview of the gap between promise and physiology (external link).
But the pattern was unmistakable: the devices on the low‑Specificity end were the ones that detect sleep well but miss wake, the kind that declare entire nights as sleep. Big no, then. And the ones on the high‑Specificity end? Better, but still far from ideal. The real prize would be high on both axes. And only one device 4 dared to sit there: Oura Ring 3. One lonely dot breaking away from the herd — a single, stubborn outlier defying the entire diagonal of mediocrity. And that was the moment I added “Ring” to the heading of my paper. A small gesture, but the kind that shifts the entire story.
Watch / Band / Ring
The second graph went further, introducing sleep stages: light, deep, and REM. On the X‑axis, the Average Agreement over 4 Stages (Awake being the fourth, by the way) — the percentage of times the device matched the heavy scientific equipment (capped at 80%, unfortunately). On the Y‑axis, the Minimum Agreement of the 4 Stages — the worst percentage recorded for any stage (capped at 70%). A good device would sit in the top right: high on both.
It’s important to know which devices are good and which are not so reliable, since some of them are actually quite bad.The Quantified Scientist
So, in one video, I would have all the devices (at the time it was released: 2024). All the data. All at once. Not one‑by‑one reviews. Not isolated anecdotes. But a full, panoramic, scientifically grounded comparison. The perfect way to know who overestimates what, who underestimates everything, and who pretends you slept like a baby when you actually spent half the night negotiating with your pillow.
One brand after another marched across the scatter plot like soldiers in a doomed parade, each one stepping forward only to be executed by my pen. I was waiting for the firmament, the top‑right corner of the graph where the good devices lived, the ones that didn’t confuse sleep with unconsciousness or wakefulness with existential despair. Until then, everything else was just another dot to eliminate, another pretender to cross off. Xiaomi appeared first — a timid little dot clinging to the wrong end of the graph — and I didn’t hesitate. NO. I wrote the name only so I could scratch it out with the satisfaction of finality. Garmin followed, and met the same fate: NO. Huawei: NO. Another clean strike. Polar drifted into view next. NO. Then Galaxy — a name that promised the heavens but landed somewhere in the statistical mud. NO. And so on…
And then, finally, the parade changed. The diagonal of disappointment gave way to something else: the good devices. Fitbit — a ghost from 2015 — appeared first, not as a single dot, but as a constellation of points scattered across the graph, all sharing the same algorithm, all whispering the same cautious promise. Not perfect, not transcendent, but undeniably better than the others. So I wrote it down. Then came Whoop — a strap I had never even heard of. A “maybe” formed in my mind, the kind of maybe that could have grown into something more, and I placed the name gently inside brackets on my paper, as if protecting it from premature judgment. And then came Oura. Not one dot, but several — generations, iterations, evolutions — but one in particular stood apart in the top‑right corner, glowing like a tiny scientific revelation (still capping at 80 %, though): Oura Ring 3, crowned by the data as the best sleep tracker out there.
Even better than the Apple Watch 8.
He then showed his own data — his personal EEG/PSG testing — and remarkably, it matched the literature almost point for point. Except for one twist: in his hands, the Oura Ring 3 and Apple Watch 8 swapped places. But I didn’t care. A watch and Apple? A double no. The rest of the patterns held, and that was enough for me. By the end of this scientific tour de force, the list had narrowed itself to three 6:
Fitbit, Whoop, Oura
A strap. I had never even heard of that. Should I add a new heading to my piece of paper? So when Whoop appeared on the graph — confidently, even impressively — I stared at the screen with the same bewilderment as someone discovering that phones no longer have buttons. But I was curious; curious, but cautious. Remember, I placed the name inside brackets on my paper, the universal symbol for “I don’t know what this is, but let’s pretend I’m open‑minded.” Now, my curiosity didn’t last long. Then came the fatal words: “monthly fee.” A subscription. That criterion was one of my lonely three on the list: waterproof, no subscription, long battery. And the “no subscription” part was non‑negotiable. And since Whoop wasn’t even among the top‑ranked devices anyway, I didn’t think twice. So the bracketed maybe became a clean, merciless strike, and with that, my list shrank to two names:
Fitbit, Oura
But my mind was blind to Fitbit. It was all about the ring, the best sleep tracker out there.
And I started imagining it — not as a gadget, but as an object. I pictured the heavy, ostentatious rings worn by US sports champions, the kind that look like they could double as blunt weapons. I pictured the signet rings of old Roman generals (at least the ones Hollywood gives us), carved with eagles and laurel wreaths. I pictured Aragorn’s ring — the quiet authority of it, the weight of purpose… And so, as the weekend stretched ahead of me — a weekend I had sworn would be for rest, for my son, for breathing — I found myself leaning closer to the screen, eyes narrowing, pen in hand.
To be continued…
1 The video was published in 2024, so of course, not every shiny new device made the cut. And scientists, being human, tend to study the brands everyone already knows — the ones with money, marketing, and research budgets. ^
2 A PPG sensor is an optical measurement device used to detect changes in blood volume within the tissue’s microvascular bed. It is the foundational technology behind the glowing lights on the back of smartwatches, fitness trackers, and clinical pulse oximeters. The sensor functions on a simple principle: blood absorbs light differently than surrounding tissue. Every time your heart beats, a pulse wave of blood pumps through your blood vessels, temporarily increasing the blood volume in your capillaries. By reading the frequency and structural shape of the optical pulse waves, PPG sensors extract broad physiological data: Heart Rate (HR), Blood Oxygen Saturation (SpO₂), Heart Rate Variability (HRV). ^
3 CuriousCatalyst (2025) Sleep Tracking: Why Even the Best Wearables Are Only 65% Accurate (external link). Medium. ^
4 Not entirely alone — the Apple Watch 8 was there too. But as The Quantified Scientist pointed out, both devices owed their lofty position to studies sponsored by Oura and Apple. Not exactly the purest form of scientific independence. ^
5 Tracking these stages requires brain waves, eye movements, muscle activity, and more (see textbox). What a sleep lab does with a forest of electrodes can only be approximated by a wearable. ^
6 He also mentioned the EightSleepPod, but this one never made it onto my list. I hesitated for a second when he said it was the most expensive device in the entire overview — curiosity is a stubborn thing — but the moment I realised it wasn’t a wearable at all, but a mattress pad, it was game over. Not exactly something I could wear to track my steps. ^