I’ve been wearing a Garmin on one wrist and a Fitbit (Pixel) on the other. Here’s how they estimate my calories burned:
The dotted gray line is what you get if you multiply Garmin’s numbers by 1.749, which is (I think, I’m trusting ChatGPT on this) the mean-squared-error-minimizing multiplier.
I noticed last year when I switched to Garmin from Pixel that I was apparently burning fewer carolies in my workouts. Since these are close to identical workouts, I presumed a difference in algorithms, and did a bit of looking into how the underlying algorithms were built.
Turns out they were based on a small number of papers, generally looking at young, moderately fit, adults, doing moderate intensity workouts - for example, this one: Prediction of energy expenditure from heart rate monitoring during submaximal exercise - PubMed , looking at “Regularly exercising individuals (n = 115; age 18-45 years, body mass 47-120 kg)”, working at “three steady-state exercise stages on either the treadmill (10 min) or the cycle ergometer (15 min) at 35%, 62% and 80% of VO2max”.
The resulting models, given they started with n~100, have a bunch of degrees of freedom that are poorly controlled. I’m outside of the age range, doing workouts that are longer and intermittently more intense. So no wonder they give different numbers - all I can usefully do is hope they’re directionally correct, and track intensity but not necessarily carolies.
I do sometimes wonder if they are susceptible to the distortion you get in clothing sizes, where some shops are famously generous in their use of (in particular women’s) clothes sizes. Wouldn’t you want a workout device that said you were working harder??
I experience the same thing (also wearing both kinds of watches 24/7). From my net calories goal I can say for my case Fitbit was way over counting. Even with targeting a daily 1000 calories deficit with Fitbit I would still gain weight but with Garmin I lost weight as expected.
I’m still waiting for Garmin calorie integration but for now I just have a manual goal.
Ah, you mean getting active calories as an autodata metric?
I agree, that’s worth doing.
One frustrating thing is that Fitbit (or the Pixel watch in particular) is much more accurate than Garmin for heart rate. I know this, ironically, because when I use the gold standard – a Garmin chest strap – my Pixel watch matches it almost perfectly. My Garmin watch on the other hand, is all over the place. That may or may not matter much for calorie estimation but it’s not exactly reassuring.
Also, I’m realizing a problem with the graph at the top of this thread: I should exclude the spikes. That’s because those are days when I biked and in those cases my Garmin is getting exact calorie numbers from the power meter in my pedals. So not an apples-to-apples comparison in terms of computing an offset or multiplier to translate between Garmin and Fitbit calorie estimates.
Ooh, interesting! What’s your species of Garmin watch, if I may ask? Mine seems pretty accurate for heartrate, compared with chest strap (I get to compare them on the bike, which talks to my chest strap). My watch is a Garmin Descent G1, for comparison.
It’s a Garmin Fenix 7 Solar, acquired 2022-04-18.
I’ve further tested my Pixel 3 watch and it’s pretty spot on compared to the chest strap, both when working out and at rest. My Garmin watch is pretty wildly off. Takes a long time to notice my heart rate is elevated, and sometimes shows it elevated for a while after it’s gone down.
I definitely agree that my Garmin watch (epix pro 2) is worse than pixel 2 for sudden changes but seems to agree with my pixel watch at the steady state. I usually don’t bother with the strap for shorter steady state activities.
It also has weird problems depending on the weather and perhaps how much I’ve sweat…
That said even for DDR sessions it does well enough that I don’t worry too much about it. (Directionally correct)
