Apple says Series 12 has the most accurate heart rate sensor. The study has caveats

Apple spent a chunk of its September 9 event on a sentence that is unusually easy to check: "Apple Watch has the most accurate heart rate sensing in a wearable." Not the most beautiful, not the most intelligent. The most accurate. Apple then published a study to back it, which is where things get interesting.

The Series 12 and Ultra 4 share what Apple calls a new Health Sensing System, and the physical changes are specific. Apple says the ECG electrode surface area grew within the same case footprint for better skin contact, the photodiodes were redesigned into a ring to catch light more efficiently, and the green LEDs are larger and more power-efficient.

Ray Maker of DC Rainmaker, who has been picking apart sports watches for close to two decades, put numbers and layout on that. He described the electrode surface area increasing by 24 percent, the photodiode array going from four pockets to eight windows, and the LEDs being clustered by colour into four green, four red and four infrared. That last change matters more than it sounds. Background readings used to lean on the infrared sensor. Now they run on green.

The payoff is sampling rate. Background heart rate moves from every five minutes to every five seconds, which Apple counts as 60 times more frequent. Heart rate variability now samples as often as every five minutes, 24 times more often than before. Workout sampling is unchanged, because it was already ramped up.

Rob ter Horst of The Quantified Scientist, a postdoctoral researcher who tests wearables against a Polar H10 chest strap for a living, read Apple's paper and found the setup close to his own method. Apple enrolled roughly 1,460 people, with just over 1,200 contributing data to the final analysis. Participants wore a Series 12 on one wrist and a competitor on the other, randomised left and right, with a Polar H10 as the reference. Exercises included outdoor running, treadmill intervals, cycling, HIIT and strength training, plus background readings during rest, desk work, meal prep and walking.

The competitors were named: Garmin Forerunner 970, Pixel Watch 4, Huawei Watch 5, Galaxy Watch 8, Whoop 5.0 and Oura Ring 5. Apple won or tied everywhere. The part ter Horst found more revealing was second place.

Across running, cycling, HIIT and strength work, the Pixel Watch 4 landed second and the Huawei Watch 5 third, with the Garmin Forerunner 970 and Whoop 5.0 behind them and the Oura Ring last. His reading of why: the smartwatch companies that poured money into machine learning have overtaken the endurance-sports brands at pulling a clean pulse off a moving wrist. A ring, he added, is simply a poor form factor for the job.

His verdict on the methodology was not grudging. Based on the published methods, ter Horst considered it a thorough study and accepted that Apple Watch led under Apple's calculation.

Both reviewers arrived at the same caution from different directions. Maker was blunt about the conflict inherent in a company sponsoring and publishing a study that favours its own product. Ter Horst likewise advised caution, while crediting Apple for publishing a result that puts a direct rival second.

His sharper technical worry is dropouts. As he described it, the previous Apple algorithm recognised a bad signal and discarded those values, which is not how most wearables behave. If Series 12 still drops low-confidence samples while competitors report them anyway, an error-based metric flatters the watch that stayed quiet. Apple's paper, in his reading, does not settle that. He also flagged that some competitor data arrived over Bluetooth broadcast, and it is unclear whether timestamps travel with the measurement or get stamped on receipt, which would introduce lag.

All that sampling exists to power Readiness, Apple's late entry into a category Whoop, Oura and Garmin have owned for years. Each morning it produces a score from zero to ten mapped to four calls: Recover, Pace yourself, Ready, Go for it. It draws on activity, workouts, training load, overnight vitals and sleep against your baseline, and it updates during the day. It requires a Series 12 or Ultra 4, because it needs the new sensor cadence. Daytime HRV, exposed as a readable number roughly two hours after you wake, is genuinely unusual among mainstream wearables.

The other half of the launch is Audio Intelligence, built on the S11 chip and its Secure Exclave, including a rolling 15-second Live Rewind transcript and Siri Recap, which ambiently summarises conversations without storing audio or attributing speakers. Maker's reaction was ambivalent rather than hostile. He can see the meeting-notes case, and still said he does not want his watch summarising every conversation with his wife and kids.

Pricing held: Series 12 at $399, Ultra 4 at $799, both shipping September 18. Ultra 4 gets past two days of general use, up to 18 hours of full-GPS workout tracking, and up to 45 hours in the new Max Extended mode. Readiness and the accuracy claim ship on day one. Independent verification does not, and both reviewers have promised their own numbers.