Is Epix Pro (Gen 2) the best watch for sleep analysis?

Well, it appears there is a lot of opinion out there and a little angst about 'The Quantified Scientist'. Maybe I should just stop sleeping. That might solve the issue :-)

That's because there are a lot more than 216k idiots in the world.

Sleep tracking is based on movement and optical HR - optical HR works differently for everyone and its not guarantee as to how well it will do with you. I would personally not suggest buying any one watch purely for sleeping. If you want the most accurate device purely for tracking sleeping do yourself a favour and by the withings sleep mat: https://www.withings.com/uk/en/sleep-analyzer  Depending on your age group a series of french hosiptals that did a fairly large trial on them found that for most age groups it was about 75% accurate (note that for some age groups it was less than 60% - if I remember correctly it started to fall off if you were over 55).

Re a watch: buy it based on your other requirements first. But how does garmin sleep tracking (first beat not the old version that is running on F5) do - 1) don't bother looking at the sleep stage stuff - instead focus on the sleep score breakdown; 2) Sleep score breakdown includes a breakdown of a) duration, b) stress, c) deep, d) light, e) REM, f) awake/restlessness. I would say that based on my scores that the main component on your score is b) stress, then a) duration, then c) deep, then f) awake/restlesness - not even sure that d) light & e) REM really factor into the score at al (if they do its probably less than 3% for both).

The good news with that is:

b) Stress is great at showing changes in lifestyle/habits/illness/injury - all factors that can cause issues around sleep and adversley impact it (case point - have an alcoholic beverage within 8 hrs of sleep and see the impact on your stress score / conversley don't (not the don't blame me if you stop drinking anything in the evening) - I would focus on this main factor and try and improve (of course not possible with illness/injury - but at least you know reason for poor sleep) . Even with poor duration as long as stress is good you will still be around 60% or fair stress & fair duration be +75%.

a) duration - yes it can sometimes get it wrong - they all can - if you are really inactive and your HR drops sufficiently then yes you watch may consider that you are asleep. Conversely if you don't actively move around upon awaking - again it might consider that you are still asleep - they all have this issue. When getting up, always check sleep score widget on watch - if watch not correctly recognized wake time - you get the option to stop sleeping - so can finalize wake time. I generally find that I am within 15 mins of start & wake time of when went to bed and got up - and maybe once every two months or so I have to manually stop sleep (usually when have to get up way before scheduled wake time - so suspect this is more a case of watch thinks you might roll over and go back to sleep.

c) deep - match against results of CPAP - when I have a bad CPAP record - I have bad deep sleep. I can't remember a time I have had a good CPAP result and a poor deep sleep result. So duration and when it occurs wrong (but who cares) but sleep score result seems accurate

f) awake/restlessness - doesn't always get this right but would say that around 80% of the time it does, and if using multiple alarms it usually shows a wake point for those alarm times - about 80% of the time (if it doesn't register one its usually the first alarm).

If just looking for a watch that will track your sleep - a good option may be the new Venu 3 with naps and enhanced sleep coach.

Re QS reviews - considering all the other main reviewers all say the same thing about why they don't analyse sleep tracking (they usually give a small synopsis of what the watch offers) - the tech isn't here yet - so no point. I give him kiudos for his attempt and his belief in himself - but considering he fails to state his gold standard device isn't gold standard as only up to 80% accurate. His lack of decent size data polls to analyse (he does 3-7 data polls - if that many) and justifies his decision on that (Ray, Des, Chase the Summit, Fit Gear Hunter etc - are all using these watches for 1-3 months before putting our reviews and are basing theirs on at least 7 times more data polls). While everyone has inherit bias QS seems pretty blatent - but not sure if just language barrier (how he phrases stuff) or just is biased - and  I do wish that we would make a much better clarification'/justification as to how he rates each device - as he will rate one as good and give it 3 stars but rate another one as good and give it 4 stars - generally with no justification between his different goods as to how a good can be 3 yet also 4 stars. I have also noticed that he is generally completely out of lock step with other reviewers. Most of them tend to find pretty simliar results - however he seldom does - does make me wonder a bit - but possibly purely due to this testing methods (several of which are flawed) and size of data polls - as so few one bad result can make a massive difference - compared to the others with far more they are able to consider whether its a trend or a one off.

Hi All,

I really appreciate all the feedback. I have never had so much response to a query on any forum in all my life (That I can remember - I am old). 

Given the feedback and all the research I have done, I have determined that knowing whether any watch is going to work for me in regard to monitoring sleep is going to be a lottery. If I really want to monitor my sleep properly, I should forget about the possibility of using a watch and source a dedicated device specific to that purpose that is consistent, regardless of body type, build or whatever (if one exists). 

I am more than happy with my Garmin 5s in all other ways. I am not into all the new fandangled stuff about answering my phone and reading texts through my watch, etc. So I am sure my 5s will last me for years moving forward.

Thanks again for all your help

Regards

Alan 

I agree with what you have stated in your bullet listing - however that is not what QS is getting or suggesting.

Some of of QS testing methods are fine others have serious flaws;

1) his test poll size of 3-7 events - surely you can't consider that a reasonable test pool size. Technically the other reviewers basing theirs on 20-60 (based on 1-3 months of daily use) is barely sufficient but overall will give a much better idea of accuracny levels as are more able to distinguish between an ongoing trend, occasional hiccup or one off occurance. You cnanot do that with 3-7 events. So no his data collection is woeful and can potentially give very flawed results.

2) valid comparison - he only ever tests with a baseline i.e. chest strap & device. He never now includes additional devices - so can never be sure that a chest strap isn't having a wobble (it happens- dry contacts etc) or whether the wobble that does occur might have occured with other watch devices - note this is someting that all other reviewers do (at least he has stopped wearing 4 watches on his wrist like some of his really old reviews where it could potentially have impared results due to constricting blood flow). And I will say that at least he seems to be trying to improve his methods i.e. adding charts so that can see lag and TBH the correlation charts I am not sure they are that helpful, as if hitting peaks and troughs but just a lag you are still getting all the training beneit -its just a lag in HR.

3) sleep stages - garmin normally pretty much nails deep sleep in his reviews - its more light/rem - the reality that even his so called gold standard EEG device (which is only up to 80% accurate at its best and is often less than that- so far from gold standard - and why doesn't he reflect the reality is that it isn't gold standard and is at its max only 80% accurate - or is he concerned that it will only validate how inaccurate watches are) will struggle to accurately determine light/rem - in fact even sleep clinics struggle to differentiate and they are using higher quality, more sensors and delving into visuals clues in much more depth. So never overly worried about difference between light/rem. And why doesn't he do any clarification measurent' analysis around HRV which is far more accurate and effective than measurement than light/rem sleep especialy as its very effective in highlighting changes in habits/lifestyle/illness/injury - or is due to non fitness watches not having it and having to use a 3rd party app on apple watch and then still have to heavily massage the data to get decent readings......The reality is the most accurate consumer sleep tracking is a withings sleep mat - and that about 75% accurate for most age groups between 25-50 (and still struggles with light/rem where its far less accurate) and pretty woeful outside that group - which is why I suspect French hospitals (they commisioned and trialed the device across a 300-500 poll group over an extended period) decided not to invest installing them.

What part of my bulleted list do you see QS having a different conclusion?

Yes QS has "fewer events" but each of HIS events measure the watch against an imperfect-but-credible reference device Dreem2 EEG headband. I personally find that to be a more valid test than other reviewer wearing just the watch for 3 months (and no other measurement to compare with).

1) Dreem is only for sleep tracking - nothing to compare it to as no one else does it. Credible - if you consider something that is 80% accurate as credible - either way it should be clarified in his reviews - which he does not. 

2) His latest reviews on the Elevate 5 sensor definitely do not agree with what Ray and Des and the others have stated. if you think it does - then it suggests that a correlation co-efficient of 0.85-0.99 is neligible and of no material difference (which depending on test subjects is generally a reasonably accurate conclusion - but is not how he states it).

His tests should still have another watch at same time - a control device, especially due to his lack of test data.

It comes down to my comment re: are the variations truely reflectiive, occasional or one time.- which especially a concern with his size of his data pool (its why imperial studies use 100s to 1000s of test sujects and are often run over 3-6 months).

You have the base (Dreem for sleep - which isn't a particulary accurate  base device / the chest strap for HR which is a very accurate base device); you have the test device and you have a control device. The test device shows a variation to the base - the big questions does your control device (which should be your next best device to the dreem/chest strap - or at least a constitent good all rounder - apple watch is probabaly a good control device) show the same variation. If no then yes that may point to being something the test device cannot do - however if the test and control device thow the same issue then you get a true idea of how good the test device is. [Note this is how all truely scientific studies work - take a look at any medical study - you have the base (the target/your goal - may be another compartive medicen's result) - you have the item you are testing (usually pil/medicine), then you have the control group (usually a placebo group). You will note that the other reviewers use a base, test device and generally at least one control device. Does he use a control device - not seen one recently - occassionally does show against another device - but not taken at the same time - in which case its a reference not a control.

I've found the Fenix 7 is much improved on the 5. The 5 used to not detect the first 30-120 minutes of sleep, or it'd think me watching TV before bed was me going to sleep early. Rarely did it get the start time right. Fenix 7 is much better, almost always within half an hour and VERY rarely is it early - if at all then by less than 5 minutes.

As others say, it tends to be overly eager to add awake time to my nights. But overall I think it's fairly accurate. If you're too active in the middle of the night it'll stop the sleep and not re-start it. But you can then just manually tell it what time you actually woke up. After a while it'll re-analyze what happened between the original stop time and the new stop time.

Epix 2 is basically a Fenix 7 with AMOLED, so if there are differences between the Epix 2 and Fenix 7 I'm not aware of them. I'm just assuming my experience will be the same (and I'm reading this sub because my next watch I'll probably swap over from Fenix to Epix).

I purchased new, pretty much the day the EP2 was released, it hasn't once tracked sleep properly. My Oura Gen 3 ring, is FAR far more accurate, very disappointed.