How Apple, Huawei, Fitbit, ouraRing crushed Garmin 965 in HR and sleep tracking

All watches need to find the pulse. My guess is apple wait until this happens and then smooth the data. Either is fine in my opinion.  It has an insignificant effect on the data over the day.  

Yes and no. I have waited for the internal HR reading to stabilise/lock on and it reported ~110bpm but that felt (physically) too low so I reconnected the external HR, which returned ~135bpm. That is a big deviation from the real HR but whether that really matters is another matter. For me, that is a clear sign of the inaccuracy of the HR sensor which is disappointing because l like pretty much all other aspects of the watch.

I do not want to let you down, but these kind of "inaccuracies" (I'd say critical design flaws) are all over the place in all garmin devices.

From the training load, ETA calculation, through daily HR and stress logging to even the auto start / stop function (that affects the trip distance a lot), and not to mention the zillion bugs that never ever addressed but you will run into immediately by just starting using a device on a very basic level - they are everywhere.

Garmin devices are cool gadgets able to motivate you to keep moving.

And that is awesome.

But that's all they can provide - otherwise they are very unreliable in all important aspects (training or health tracking) and they were always like that and I cannot see signs that it could change in future.

I find the sleep tracking to be completely off. I have worn the girlfriend's Vivoactive 4s the last couple of nights and both differ completely. 

With the 965 I mostly get bad sleep scores, even though I sleep min 8 hours and feel recovered afterwards. Apparently there is not enough REM sleep (although I remember dreaming, which should mostly be allocated there). It also mixes up being awake with REM many times. Also, like shown in the youtube video, it doesn't show any logical sleep cycles (deep-light-REM cycles).

The Vivoactive 4s on the other hand, seems to show it much better and gives me much higher scores, more cycles, lots of deep and REM sleepes and depicts much better of how I would subjectively rate my sleep.

It is the same when she wears both watches, the 965 gives just poor scores to the both of us.

That's all okay, I don't expect perfection there. However, bad sleep scores influence recovery and training readiness substantially. Due to the bad scores of the 965 I would just need to rest and never do any sports again. A lot to improve there! At least, they could introduce something like a personal rating of the sleep (like you can do after running) and have that included somehow in the rating. At this point, I'd rather input my sleep scores manually than having the 965 do it.

On the contrary, the HR measurement works extremely well for me. Really spot on most of the time.

Sleep is a funny one and it’s hard to say whether one watch is more accurate than the other.  

I do share your frustration that the training readiness may not be accurate.  However, you know your body better than the watch and can decide how to train.  

I would imagine the algorithms may differ between the watches, though this is a guess. 

The training readiness is simply a gimmick. 

Calculated in an unknown way from many already correlated data, and those data are very unreliable and calculated in a very questionable way.

Moreover, probably the most important data (the RHR) is not even calculated.

I train myself on the edge of overtraining, so I really try to take attention on these data, and really, the trading readiness shows very little correlation with my physical state.

To break down it a bit: when it is very low then obviously something wrong, but you feel that. If it is very high, then again, you feel that as well.

The real problem is where it would be really important, on the middle ground: when you are okayish but not really sure how much you can bear - and then TR is just simple noise.

The best formula I could come up is check RHR (in more weight) and HRV (in less weight) considering when was the latest session: late afternoon sessions could really ruin the sleep HRV and RHR, but that does not mean that you won't be ready to exercise next afternoon.

Surprisingly, the sleep data does not really affect my real readiness. Obviously, very bad sleep does, but in the middle ground (poor or fair sleep by the watch) is not really important. Stress score in TR is a joke, if you train hard, your stress will skyrocket almost all of the time.

I am not an athlete, so in extreme load sleep could affect more, but then again, in this case this watch will tell you nothing significant about you real sleep status.

What we have yet: recovery time - I have just arrived with a day long hiking (load: 115) and my recovery time is zero by the watch - do I need to say more about the significance (or lack thereof) of recovery time?

Acute load - the training load incentives the short, high HR activities, but scores very low the long zone 2, or even exercises with variable HRs (such as all long exercise sessions). But those put even more stress on your body - affecting the TL/TR just a bit.

If you mainly focus on such kind of endurance exercises, then you will overtrain yourself by a lot if you listen to your watch.

Now, I am not talking about some nuanced differences - so it not that TR will match somebody but not for others. There are fundamental and very significant flaws in the calculation, so if anyone is fine with the TR then it is pure coincidence.

You can consider TR at this point just a random number (in best case), in worst case, you can listen to it, and it can lead to very serious problems.