Sleep tracking accuracy - latest status

I have used a 6X Pro Solar and Oura ring since December 2019. In general I would say that, whenever I make direct comparisons in data, they are mostly in agreement in the broad scheme of things. There may be differences in the details, but mostly you could not make the case that one was more correct than the other without a third, more scientific analysis.

Here is an example of my results for last night. FWIW I did eat a lot last night and finished eating quite late - around 21:00. I did also have to get up at least a couple of time for old man toilet breaks. I would say that the broad picture from both devices is close enough.

6X Pro Solar....

Oura ring....

I am using the latest beta firmware on the watch and latest production firmware on the ring. I would say that comparisons have remained consistent on any and all firmware versions for almost two years.

I would also add that the focus for Oura is very much about sleep, and it is not designed to support and measure exercise in any meaningful way. It also sets shockingly low targets for activity for people who actually work out regularly. As and when my Oura finally quits working I would not replace it. The 6X is perfectly adequate for my needs when it comes to sleep measurement and, of course, it offers so much more on top.

I am not a sophisticated consumer of sleep data and I have nothing to compare the F6 to in this regard. However, for me the F6x totally meets my needs in terms of sleep tracking. I feel it is accurate in that it generally corresponds with my sleep experiences. I don't know how accurate the fine details are, but I don't pay much attention to those. My guess is that no wrist worn sleep tracker is going to be hugely accurate with respect to these details.

Very interesting to see both replies (eezytiger and Niterider). This sounds like the level of accuracy I am looking for. Also interesting (and encouraging) to read that this has been consistent across firmware updates.

Am then still wondering where the reports of inaccuracy are coming from (and actually I perceive them as being too many to be outliers, but maybe that is because I have ben focusing on these specifically)? 

Is it that anatomic situation of wrists is so different that sometimes the sleep tracking just does not work for some people? Differences in hardware quality? ...?

3214823 said:

Am then still wondering where the reports of inaccuracy are coming from (and actually I perceive them as being too many to be outliers, but maybe that is because I have ben focusing on these specifically)? 

If you go to sleep same time every night, wake up same time every morning and sleep like dead man all night (not moving, turning...) sleep tracking works great.

Every other scenario - it depends.

Bed times/Wake times go wrong, sleep/wake phases during night go wrong... You may even experience waking up, going to toilet and going back to sleep without watch recognizing that you were awake (funny part is that you gain "Steps" during sleep, so the watch is aware that you were walking).

3214823 said:

Is it that anatomic situation of wrists is so different that sometimes the sleep tracking just does not work for some people? Differences in hardware quality? ...?

I'm not a Fenix owner, and all the Garmin's devices I had so far wasn't using FirstBeat(at least officially) used on Fenix 6 for sleep analysis. But as far as I know even after FirstBeat estimated your sleep it can get corrected during the day by Garmin algorithms, so I think what I say may be somewhat applicable for Fenix too.

Most I'd just echo what iLikePotatoes wrote, but wanted to add that Garmin's algorithms seems to be biased to the time you go to bed, as when I was sleeping roughly from midnight till 8AM there were no issues with sleep detection. But if I get to bed on a regular basis e.g. at 4AM, and I have my bed time set correctly in the profile my watch could badly fail to recognize when I sleep and start sleep tracking from 1AM or so and I have to correct it in the morning.

On the good side, lately something has happened and Garmin seems not to overestimate sleeping time as much as it used to, so that's quite encouraging and means that either they are working on their algorithms or it's training with the time if it's AI based.

You may also want to check this guy, he's reviewing Venu 2 which is Firstbeat driven just like Fenix and the results are quite encouraging www.youtube.com/watch

Well I don't have fixed sleep times and it works well in detecting if I have to work early or not, I just set the sleep time window big enough to cover both scenarios. 

I NEVER have a good sleep score.

I have to say that I have a bad/light sleep pretty often, but even when I wake up and feel very well rested and that I had a perfect night, the watch tells me "60 points, your sleep sucks".

It is very accurate with the time-tracking in when I go to bed and when I wake up, but the rest of it is (for me) totally unreliable.

The watch tells me everytime that I'm awake in the night for over 1.5 hours but in generally I just wake up 1-2 times in the night to go to toilet and maybe because I was dreaming something bad, so I'm awake at maximum 20 minutes.

Mizzu said:

I have to say that I have a bad/light sleep pretty often, but even when I wake up and feel very well rested and that I had a perfect night, the watch tells me "60 points, your sleep sucks".

The opposite for me: it often tells meI've got a good night of sleep even when I wake up very tired. However it works when comparing from night to night: most of the times when I had a terrible night of sleep it tells "fair" (or whatever is the English translation), so it's still sort of useful to me.

iLikePotatoes said:

Every other scenario - it depends.

Very good synopsis. I feel (for what that's worth) that tracking of deep and rem sleep are fine. It falls down on light sleep because it seems to have little capacity for distinguishing between lying awake quietly and light sleep. If you tend to do the former, as many older people do, it will overestimate the latter.

I have been a little concerned about the report sleep stats which constantly show I have very low deep sleep.  However recently I have seen strong evidence that the algorithm used to detect sleep (which would presumably include sleep stages) is flawed or at best unreliable.  Last night I watched television until around 11pm, and then retired to bed.  In the morning, my Garmin reported I had gone to sleep at 7:30pm and at around 8pm was in a deep sleep for an hour.  I've noticed these kind of weird results recently and wonder if I have something wrongly configured, or not configured. Another case a week or so reported me sleeping when I was driving.

I previously used a Fitbit for sleep tracking which seemed to be very much more accurate, and representative of what I would have expected. 

Would be great to know if Garmin are planning to improve the algorithm, so that at least it looks like something reliable.  Or even better if there is something I can be attributed to the bad reports that can be adjusted.

Tim.