One tester's EEG data put Oura Ring 5 behind Ring 4 on sleep stages

Rob ter Horst spent more than a month sleeping in an EEG headband so he could check whether the Oura Ring 5 knows what your brain is doing. His result on The Quantified Scientist was awkward for the new model: on his body, its sleep-stage readings agreed less often with the EEG reference than the Ring 4 did.

Ter Horst, a postdoctoral scientist working in biological data analysis, ran the Ring 5 against a Zmax EEG headband — a device he describes as "sort of a silver scientific standard, not the absolute gold standard, but good enough to give us an impression." Rather than pick representative nights, he pooled every usable measurement.

Deep sleep held up well, agreeing with the headband about 81 per cent of the time. Light sleep landed around 72 per cent. REM sleep is where it came apart, at roughly 57 per cent — and the way it failed matters more than the number. Chunks the headband scored as REM came back from the ring as awake time, with the rest spilling into light sleep. Position was mostly right; the labels were not. "Your sleep cycles you would be able to see also based on your data from the Oura Ring," he noted — a narrower claim than the app's stage breakdown implies.

In his comparison chart the Ring 5 drops out of the top group. Ter Horst's first tier holds the Oura Ring 4, the Whoop strap, the Eight Sleep Pod, Fitbit and Google devices including the Pixel Watch, and Apple Watches. The Ring 5 now sits closer to what he calls second tier — newer Amazfit hardware, recent Garmin models, 2025 Galaxy watches. His composite sleep score for it: 70, a rating of "good but not great anymore."

Here ter Horst is scrupulous. He also computed Cohen's kappa, the more traditional sleep-staging statistic, and on that measure the Ring 4 and Ring 5 come out identical. His own score averages sensitivity across the three stages so that each one counts equally; kappa does not. As he put it, if Oura were optimising for kappa, "then yes I would get the exact same result basically" for both generations.

So a firm could tune an algorithm, watch its internal number hold steady, and still ship a ring that scores REM noticeably worse against an independent reference. He does not accuse Oura of that. He says he discussed the findings with the company and that its internal testing shows no meaningful difference between generations, and that Oura intends to submit its sleep-staging validation to a scientific journal and share the results with him once it is accepted.

He is also careful about his own limits: "My results are real, but still representative of one ring tested on my one body." He wants to repeat the work with a second ring, a different finger and a second person before calling it a final review.

Overnight heart rate and heart rate variability came back among the best results he has recorded from any consumer wearable — good enough, he reckons, to sit close to an ECG chest strap. Exercise heart rate is more mixed: correlation of 0.95 for running, 0.91 for indoor cycling, and a poor 0.72 outdoors, where city riding with its traffic-light surges left the ring tracing an average through changes it never caught. Against everything else he has tested, it beat only 13 per cent of devices indoors and 19 per cent outdoors.

Ray Maker of DC Rainmaker, who bought his ring and pays his own subscription, found workout heart rate broadly reliable, with a distinctive stumble on descents that cleared the moment he was back on flat ground. His complaint is about the product around it. The new live activity mode will not show you your heart rate at all unless you pair a separate Bluetooth sensor — a strap, a watch, a pair of AirPods Pro. And when a workout syncs to Strava, the heart rate data Oura already holds is stripped out. Maker called that "the most lazy integration ever to the biggest platform in the sports tech world." He also flagged a VO2 max estimate of 38 against a measured figure near 60, and confirmed no software feature is exclusive to the Ring 5: the changes are size and battery.

Both mainstream reviews land somewhere warmer, and for consistent reasons. CNET's reviewer, two months in, said the ring vanishes: "the best wearable isn't necessarily the one that's the most accurate. It's the one you're actually going to live with 24/7." Eight days of battery, roughly a day more than the Ring 4, means it stays on overnight while a smartwatch sits on a charger. After a few weeks of manual tagging, automatic detection started catching Pilates sessions about 90 per cent of the time. On a run against a Polar chest strap, average heart rate matched and peak was within seven beats.

Tom's Guide reached the same place from the software side, pointing to menopause insights and cycle tracking that now accounts for hormonal birth control — features shipping to older rings too — and concluding that "this is a product where maybe software even matters more than the hardware."

Those two positions are not actually in conflict with ter Horst's data. A ring you never take off produces long, unbroken trends, and its overnight heart rate and HRV look excellent. What this month of EEG data suggests is that the stage-by-stage hypnogram deserves more scrutiny. That matters because Oura made the ring 40 per cent smaller, rearranged its sensors and reduced the number of optical signal pathways while saying accuracy should remain the same. Ter Horst's planned tests with another ring, finger and participant should show whether his result can be reproduced; Oura's promised validation paper should provide a broader comparison.