Hacker News Reader

Comment by xlaacid | Hacker News Reader

None. Period. I was in human sleep research for 14 years as a technologist and proj manager. I tested dozens and dozens of devices- a mattress pad that claimed to track sleep. None of them are accurate.

to track limb movements you need EMG attached to your legs and then to a recorder. Then a sleep tech certified in scoring must review the results.

A board certified sleep doctor can get you home devices to track periodic limb movement if necessary.

Acti-watches are the only devices that you might be able to get "accurate" sleep(not limbs)tracking. They're expensive and you need to buy analysis software to view results. I would not waste your money on any publicly available device- they wont get you the data you want.

View on HN

Really?

I get it, at the margin this is probably true but for gross results?

If I drink, AutoSleep lights up like the fourth of July. If I don't, it reports every metric is better. If I drink, SnoreLab lights up. If I don't, it's fine.

Same if I exercise, it's all over the results very clearly.

I'm not claiming they can measure REM accurately but for major life factors they seem pretty damn good.

Can you say a bit about how you know this? Like are all these trackers you tested compared against polysomnography/EMG? Wouldn't everything of course be not as good as that (since we're treating that as 100% perfect?

How do you know that the certified sleep tech is reporting good results? How do you know that the results have good external validity (i.e. the results are not due to the unnatural sleep study environment with EMG attached and being forced to sleep at a certain time in an artificial bedroom?

The Fitbit looks pretty close. I don't think it's perfect, but it's directionally correct.

I feel like for a lot of self improvement frameworks it's about direction not accuracy. Like for example, body fat % machines vary greatly, but surely being closer to the 10-20% range is better than 30-40% range regardless of the method you use. Probably the same can be said for sleep.

Are you aware of modern consumer sleep-tracking headbands like Dreem and Philips SmartSleep? Considering these have an EEG, I find it hard to believe they're less accurate than an acti-watch.

Dreem has a self-run study that claims pretty great results [0], do you have any objections to their methodology?

[0]: https://www.biorxiv.org/content/10.1101/662734v1

Thanks for citing the study. Just read through the introduction and conclusion and the Dreem sounds very promising.

After reading Matthew Walker's "Why we Sleep", I currently use the deep sleep percentage metric provided by the Sleep as Android app to determine how well I sleep, and jot it down in a sleep log. The book has given me the idea that deep and uninterrupted sleep is hugely important, but obviously the app is only able to give a crude estimate.

Perhaps this headband would give me better insights into my (failings of) sleep. I hope the parent commenter replies to your message.

> to track limb movements you need EMG attached to your legs and then to a recorder

I've had this problem for many years where my legs don't relax when I sleep. So when I wake up my legs are always tired and sore. It's not serious enough to cause any major health issues but very annoying. I've visited many doctors and none were able to provide any useful information.

I was thinking of attaching EMG and getting some data for analysis. But I wouldn't even know where to start.

I don't know how good professional tracking is, but I can tell you that I definitely get useful information from my Oura ring that has lead to an improvement in my sleep. Many small things, but the obvious one is how much even a drink or two of alcohol is affecting sleep for half the night.

Obviously won't help with tracking leg movements, though. Just for a quick confirmation I'd recommend a web camera, but there won't be statistics from it.

Have you looked at the Oura ring? They claim to be the best publicly available tracker.

Reply on news.ycombinator.com