[Dave_Rosenthal]

I'm curious about the word 'bias' here. From a technical perspective, does the pulse oximetry sensor just have more variance on darker skin (presumably from it being more opaque and signal to noise being worse), or is there an actual bias (e.g. reads high all the time)?

[hedgehog]

Lower overall return of light which causes low SNR, how that manifests is a software question. I've worked on similar systems and read some of the papers about the pulse ox problems, the problem is pretty well known for the finger clip style devices. Watches are harder because the fit to wrist is unreliable and there is muscle and bone right under the skin. In my experience (pretty pale skin) the data quality from the Apple Watch is poor but good enough for rough trending over weeks.