While I know that sleep tracking (especially detecting the correct sleep phases) is far from being perfect, I'm getting more and more fed up with how badly it actually works and how it impacts other metrics on the watch, in particular overnight HRV readings, the training readiness and quite possibly also the suggested workouts.
Three examples:
Today I got a really bad sleep score reading of 51. The watch detected quite accurately my bedtime around 9 pm yesterday. But around 1 am the watch stopped recording any metrics for a while (maybe it applied an update?) and consequentially detected this as my wake-up time as soon as it stopped recording. I actually slept normally until 5 am and woke up very well rested.
A few weeks ago I had quite a bad night, when I was woken up around 2 am and couldn't get back to sleep due to external circumstances. Nevertheless, my watch recorded sleep until my normal wake up time at 5 am, even resulting in a "Good" sleep quality, although I was really tired. It even detected REM phases while I was awake.
I encountered another issue on a recent weekend, when I was watching TV on the couch with my girlfriend starting around 8 pm. I was very relaxed and my body was not particularly stressed from training that day. We watched until midnight and then went to sleep. The next day I was surprised that the watch measured my sleep times starting around 9 pm. I'm surprised how this could be possible, since I was also constantly moving slightly, was on the phone and drinking water. I can just assume that it started to detect sleep because of low stress and my heartrate went quite low (below 50 bpm) at some point for an extended periode of time. The wierdest thing was that it even detected REM and deep sleep phases while I was actually awake. The worst thing again, because it thought I slept for a really long time that night (about 11 hours) it again gave me a poor sleep score. In this case I was lucky though, because after adjusting the bedtime hour to cut away sleep time, the remaining deep and REM phases were sufficient for an excellent sleep score... on bar with how I actually felt that morning.
So, I know that I can adjust the sleep time within the Garmin Connect app, but this doesn't change the sleep score significantly (today only from 51 -> 61) because it automatically categorizes the added sleep time as light sleep. But for a good sleep score you need a sufficient amount of deep and REM sleep.
It also doesn't help with overnight HRV readings, because the added time is not taken into consideration. Since I get usually higher HRV readings in the second half of the night (especially after a workout in the evening) due to getting more rested, my overnight HRV average is cut short if wake-up times are detected too early. That suggests that my body is in a more stressed condition than it actually is.
If it would only be a metric on its own, I would just shrugg it off. What concerns me the most about these issues is, that they also influence other measurements like the HRV readings (as mentioned above), training readiness and I assume the suggested workouts, too. When I get a bad sleep score like today it increases my recovery time on the watch and consequently my training rediness. Furthermore it seems to impact the suggested workout feature. It got rid of a an anaerobic session it had planned previously in order for a base session. Also, I noticed that it adjusts the Watt values for base sessions based on my recovery status. I have an cycling FTP of about 225 (manually set to my watch). Other platforms like Xert even suggest a higher FTP of 235, Strava suggests on the power curve 226, Intervals.icu gives 223. Based on my training based on FTP, I always feel appropriately challenged, therefore I think that the value set on my watch is currently ok. When I'm well rested according to my watch, it suggests Base training with 150 Watts which is well at the center of zone 2 at about 55% – 75% FTP. When it thinks I need more recovery time it sets the suggested Watts significantly lower to 120-130 Watts for base(!) training, which is already within or very close to the border of active recovery. This doesn't make any sense... if it thinks that I should only do active recovery, it should suggest active recovery since it is a possible and available workout option. But ignoring this issue, it should not adjust the training so significantly based on bad sleep score readings. In general I'm quite happy how the watch naturally suggests a lot of base training and 1-2 threshold and VO2max sessions per week, quite similar to a polarized training plan that I try to apply.
In my opinion there are very simple fixes to this problem, so that the user can intervene if the detected sleep metrics go wrong.
My suggestions how to improve these problems:
- Most importantly: Let only sleep metrics impact other functionality of the watch that the user can influence, i.e. in the current state only the length of sleep not the quality.
- If you want to keep a metric describing the sleep quality, give the user the option to at least overwrite the sleep quality based on your default categories (Poor, Fair, Good, Excellent). Then set the sleep score to an average value in the respective range. Basing the sleep quality on the sleep phase detection which you obviously can't do properly and probably never will, because it is even difficult in a lab, is just ridiculous. Why don't you just ask during the morning report how the user's sleep felt and how well rested they feel? Merge this to the actually sleep length and you will end up with a more reliable sleep score.
- You probably already track HRV constantly as a metric to determine stress throughout the day. If someone changes the start/end of their sleep time fill the overnight HRV readings with those values.
- If an update is applied during the night (sleep times), just set a flag on the watch so that it doesn't forget after booting up again, that the wearer was at sleep at the time of the update and didn't wake up.
- If it was an update tonight that messed up my sleep score, don't apply updates during sleep times automatically and let me manually apply them during the day as I do so often already.
As further information, I wear my watch daily, tightly fit on my left arm, slightly above the wrist and clean the sensors on a daily basis with water especially after training.
I have really big issues with how strongly these sleep readings affect other metrics and features on an already quite expensive watch, rendering many of said metrics/features completely unreliable and useless. Currently it just seems like a feature implemented by people, who don't know what they are doing, overengineering the feature to make it appear well thought through but eventually making the metric significantly worse than it could be, while also negatively impacting other features.
All I can say as a conclusion is that based on these issues that render so many features of the watch useless, I currently can not recommend a Garmin watch to anyone, which is sad, because in general I like the watch if there weren't the sleep scores which affect so many other features negatively.
