News for hackers. Stuff that matters.
Re:Fitbit's 150B hours of heart data by wufufufu on Wednesday September 05, 2018 @10:34AM (#17918310)
Anecdotally, the heart rate data they have isn't very good. I had a Fitbit to track my HR while running, and the HR reading would change dramatically based on what I set my stride length to in the app. I would run on the treadmill, and it would guess how much I ran. If my stride length was set to a longer length, it thought I was an elite athlete and told me my resting HR was sub 50 BPM. If I set a normal stride length, my resting HR would be 60+. That's when I stopped caring about health trackers.
Re:Fitbit's 150B hours of heart data by hammock on Wednesday September 05, 2018 @10:46AM (#17918430)
Guess what the digital bathroom scales are just as bad when it comes to body fat measurements, and even worse when it comes to weight measurements.

(The algorithm that shows your weight has hysteresis built in, so that if you step on, get a reading and step off, then step back on with a 1- or 2-lb weight, it will give you EXACTLY the same reading. This is to fool you into thinking its accurate)

Re:Fitbit's 150B hours of heart data by noarchy on Wednesday September 05, 2018 @10:42AM (#17918393)
For me, doing hill sprints on a bicycle sometimes completely fails to register on the Fitbit. My heart rate will sometimes show up in the 90-100 range on the Fitbit, despite my Garmin (which gets data from a chest strap monitor) showing me 160-170, which is far more believable.

And then there are weird situations while merely walking around where I'll see strange spikes (160+) on the Fitbit when I know my heart rate is near 90 or so.

Re:Fitbit's 150B hours of heart data by thr0w__4w4y on Sunday September 09, 2018 @01:43PM (#17946919)
I've seen the same thing. I'm a middle aged guy, endurance athlete, was competitive nationally at the NCAA level way-back-when. I learned a lot about nutrition and physiology as part of my sport.

Back then the conventional wisdom (not universally accepted) was that max heart rate was "220 minus age". Well, if you do that math on me, the peak -- not sustained -- rate that I reach now is outrageous, in a good way. Granted, I've learned to push through and perform at high levels of exertion and exhaustion, it didn't happen from one day to the next.

Anyway, there were times ~3 years ago where I'd be tearing up a workout, Fitbit would read 130, I knew it was f*Ing wrong. Bought a chest strap and saw what was really going on. Low 190s. Thanks FitBit!! I would have been happier if the FitBit just put up "??? WHAT THE HELL ARE YOU DOING OLD MAN???"

Re:Fitbit's 150B hours of heart data by Matthias247 on Wednesday September 05, 2018 @02:56PM (#17920895)
I have a Versa. I have the feeling for casual daily activity or light exercises (hiking, walking around) it's OK. The resting heart rate also seems to be mostly accurate for me.

However for more intensive activities it seem like it can't keep up. I guess either because of sweat, or because of delays during processing. When I went biking with it and going full-out, the heart rate stayed really low for minutes, until it went somewhere in the medium range (still lower than it actually was). When doing something short and intense like burpees, it might even keep displaying the low starting heart rate, even though the actual heart rate is on the highest side. My garmin with chest strap does a way better job for these kind of activities.

However in total I still like the Versa and would buy it again.

Re:Fitbit's 150B hours of heart data by phnk on Friday September 07, 2018 @05:28AM (#17932933)
> the heart rate data they have isn't very good

Not too many people seem to care on the reading end, so I guess that's fine.

As another user (kaustyap) posted further below:

> It kind of sounds like marketing pitch by fitbit to let potential clients in health industry know that they have so much of data to sell.

Sounds right to me -- it's a marketing ploy.

In an academic or industrial setting, the data might still be useful, but it would have to get re-crunched with many, many more controls, plus sample corrections, cross-validation against other data, etc. -- than what the linked article does.