I do a lot of vigorous exercise and (originally) bought the fitbit charge 3 not only to track all day activity/HR, but also during exercise. As above I found it to be very inaccurate, basically not working above 140/150 bpm. I returned it after around 30 days of frustration and frequent interaction with the support teams.
I bought an Apple watch subsequently (series 4) and have been using it since (almost a year). I also use a polar chest strap (H7) sometimes, so can compare the two - and was doing so before I got either wrist based HRM - in fact I've been working out with chest based HRMs for over 20 years (mostly polar). The Apple watch is pretty good, even during vigorous exercise, for example I do 2x10 min HIIT sessions on a rowing machine 3 times a week where my HR will range from 155-185 once 'up to speed'. 9 times out of 10 the Apple watch will pick up the HR within 15 seconds, be consistent with the strap readings to within 1-3bpm and react quickly to the changes when HR goes down in the lower intensity phases.
When I go for a run (these also tend to be high HR, 100% effort runs, not 70-80% training runs) I find it can over read the HR vs the strap by up to 5bpm and also has bias to read high (if I assume chest strap to be 100% correct).
The downside is that sometimes it can take a while to start getting an HR reading, for example on a 5k run I did in the last week the first 6-7 mins had only a couple of readings, almost like it's struggling to pick it up, but after then it did give readings (every 5 secs or so) which seemed accurate with the caveats above - slightly high. I do find once it starts reading it carries on, if it has trouble at the start, it 'gives up' and tries again later, maybe too much vibration, wrong position on wrist, something else, not sure and haven't tried exhaustively to identify/eliminate as I have a backup with the strap.
So it's a bit annoying for that 1 in 10 times when it doesn't get a reading for a chunk of the time (or not at all), but it also seems sport dependent - swimming (which is notoriously hard for optical wrist based HRMs) is much lower chance of getting a good reading throughout, I'd say 3-4 in 10 times it fails, but on the rowing machine seems almost perfect (and I have a lot of data points).
Obviously latest apple watch is different price bracket than fitbit charge, but for me it was important to be accurate at high HRs so the fitbit was a waste of money for someone who is exercise focused and wants to monitor their HR. Previous models of the apple watch though are not a million miles away in price point and are definitely worth considering. This is also dependent on being in the iPhone ecosystem, if you are on android I'd probably argue there is no point looking at apple watch. Obviously I can't comment on the android watches out there.
Hope this helps.
