Product1 publisher2 min readPublished
Chest-strap test ranks the Series 12 heart sensor strongest in cycling and weakest in lifting
Rob ter Horst put Apple's rebuilt sensor against a Polar H10 across cycling, running and weightlifting. The three correlations span 0.97 to 1.0. The every-five-seconds resting stream sat outside the test.
The Product Desk · Product desk
What happened
- Apple rebuilt the heart sensor in the Watch Series 12 so that it samples heart rate every five seconds even when the wearer is not active.
- Rob ter Horst of The Quantified Scientist ran the watch against a Polar H10 ECG chest strap across three sessions: indoor cycling, running and weightlifting.
- In indoor cycling the watch's readings were nearly identical to the chest strap's, a near-perfect correlation of 1.0.
- Running returned a 0.98 correlation with the strap, with ter Horst noting minor deviations at the higher heart rates.
- Weightlifting, which ter Horst calls traditionally one of the hardest activities for a wrist sensor, produced the lowest correlation of the three sessions at 0.97.
Compiled by The Product DeskSomething wrong?How this is made
Why it matters
- constraint The evidence available to anyone building on this sensor covers exercise only, so a feature that reads heart rate while the user sits at a desk is unverified against an ECG reference.
- decision A team whose feature fires on an absolute heart rate during strength work now has to pick a tolerance band or a different trigger, because lifting is where the watch and the strap agreed least.
- precedent A correlation per activity, plus a percentile against other wrist devices, is how the next sensor claim will get checked.
Apple's case for the rebuilt sensor is the reading you get when you are not exercising. 9to5Mac's own review describes the payoff as insight into hauling groceries, doing yard work, an afternoon cup of coffee, a stressful moment at work, all sampled every five seconds [8][1]. Rob ter Horst's comparison against a Polar H10 happened during exercise [2].
Across his three sessions the correlations span 0.03, cycling at the top and lifting at the bottom [1]. A correlation at that level says the watch trace and the strap trace moved together. It does not say the watch printed the same number, and ter Horst's caveat falls at the top of the running range, which is exactly where a training feature would be reading [4]. On lifting he said the result was "quite good" compared with the competition [7]. He ranked the Series 12 above 85% of the devices he has tested for weightlifting [6]. By that ranking, roughly 15% of them did as well or better [2].
Against his longer-term testing of the Series 11, ter Horst said the Series 12 delivered broadly similar heart rate accuracy overall, and the older watch actually did slightly better during running [9]. He allowed that Apple may still be tweaking its algorithms for the new sensor [10]. He also hand-counted roughly 5,000 steps and found the new watch spot-on or only slightly over [11]. The write-up does not disclose session lengths or how many sessions he ran.
For a team shipping a feature on top of wrist heart rate, two axes sort the features. One is whether the feature triggers on an absolute heart rate or on a trend. The other is whether it fires during strength work or at rest. Absolute thresholds during lifting sit on his weakest session, and at 0.97 against an ECG reference that supports a session-level summary better than a set-level trigger [5]. Trend features during steady cardio sit on the cycling result, where the watch and the strap were nearly identical [3]. The five-second sampling is built for the at-rest quadrant, and ter Horst's strap ran only during workouts [1][2].
What to watch
- Longer-term Series 12 testing from ter Horst. That testing would show whether the running number moves after Apple tunes the new sensor's algorithms.
- Any independent comparison that samples the resting, every-five-seconds stream against an ECG reference instead of during workouts.
- Whether other reviewers reproduce the 0.97 lifting correlation with heavier loads and longer sessions.