Dr. Michael Grandner

Exactly.

In the evening, you don't want it to.

Yeah.

Yes.

Excellent question.

People need to know that using wrist-based movement to estimate whether someone was asleep or awake across a night has been around since the 1970s.

That data has been well worked out.

Those algorithms are pretty robust.

It's actually shockingly good.

You can predict with about 90% accuracy using movement alone.

And this was analog devices that were on a tape backup or eventually 64 kilobytes of memory on the whole watch.

With that level of technology, you could get over 90% accuracy minute to minute.

Were you awake or were you probably asleep relative to brainwave activity?

Sleep versus wake.

That's what these devices are best at.

They've always been best at.

That is the data that I would trust the most.

With the asterisk of...

It's going to be different from the brainwave activity in that the brainwave activity will pick up lots of little awakenings that the movement detection probably won't.

So it will underestimate wake time relative to looking at brainwave activity.