Recently bought a Garmin 935. Loving it so far. I see where it gives a VO2 max. Have mostly used it for running. Anyone know how accurate it is?
I don’t know how accurate it is. In that it cannot monitor how much oxygen you take in or expel, it is obviously an estimate. Upon what do they base the estimate? I don’t know.
The numbers it gives me are flattering. But do they really mean anything? I doubt it.
For the record, the FTP number my Garmin gives me (auto calculate, but not based on a test) are off by about 20% (low).
If you dig into Garmin’s website/literature, they describe (qualitatively) the algorithm that is used. In essence, it is based on assuming a fixed running economy for everyone, then extrapolating from estimated submaximal VO2 to VO2max based on heart rate variaiblity (the closer you get to VO2max, the more regular your heart rate, due to withdrawal of parasympathetic tone).
If I had to guess, based on just the approach used I would say that the standard error of the estimate is probably around 5-10%, i.e., in the right ballpark, but not especially accurate.
IIRC, they use a slightly different approach for cycling, which does not rely on heart rate/heart rate variability data, but still assumes a fixed cycling economy (and a fixed relationship between 20 min power and power at VO2max? Not certain). That estimate is probably good to w/in 5% or so.
If you dig into Garmin’s website/literature, they describe (qualitatively) the algorithm that is used. In essence, it is based on assuming a fixed running economy for everyone, then extrapolating from estimated submaximal VO2 to VO2max based on heart rate variaiblity (the closer you get to VO2max, the more regular your heart rate, due to withdrawal of parasympathetic tone).
If I had to guess, based on just the approach used I would say that the standard error of the estimate is probably around 5-10%, i.e., in the right ballpark, but not especially accurate.
IIRC, they use a slightly different approach for cycling, which does not rely on heart rate/heart rate variability data, but still assumes a fixed cycling economy (and a fixed relationship between 20 min power and power at VO2max? Not certain). That estimate is probably good to w/in 5% or so.
Thanks for breaking this down. I knew it probably was not super accurate, but having a number to work on is nice. Same with the Garmin running dynamics. I imagine the numbers aren’t very useful, but anything to keep me motivated is nice.
I doubted the accuracy of the VO2 max (cycling) on the Garmin. I did a lab test for my cycling VO2 max, which was 1 ml/kg/min higher than the Garmin’s estimate. So, it was pretty dang close.
About as accurate as using the 220-your age formula to determine max heart rate (aka accurate in about 5% of the population)… that being said, while there are some faulty assumptions, it’s probably one of the less inaccurate algorithmic indirect estimations of VO2max… But that will never rival having a proper gas collection test done…
Mine is 10 points off (low) from the lab testing, so not even close. Being a tall and big guy (6ft9, 208cm) totally out of the ordinary, these estimates don’t seem to match at all with the VO2 I can really take in. But then of course, how could it account for the 1% of the tallest, smallest or whatever people. The more ordinary/average you are, the better the results will be. That’s not only the VO2 estimate, also the Race predictor, FTP test (apart from doing a real test), and not only Garmin, but Polar and others. If you want something real, do a real physiological test with a certified doctor.
It’s a bunch of crap marketed in a way to differentiate it from previous Garmin devices in order to extract dollars from your wallet.
I’m a 40 year old man who has a max heart rate in the area of 180 ± a couple beats. My resting is 32. It currently says my vO2max is 67, that I could run a 15:29 5k or a 2:28:32 marathon. My PR’s are 16:59 and 2:39:56 and are 8 years old.
Do the math. It’s crap.
For me, using Fenix 3 non-hr (I tried to use HR belt during 5 months) the vo2max results were approx 10% low.
Vo2max started at 54 and went to 60 (both on my run and bike activity), but test lab gave 67…
Also worth noting that for those that have darker skin, the optical HR tends to struggle (they need to use infrared, not the green lights to improve accuracy). I have a 235, and recently put it into broadcast mode for a Zwift ride, and for Zone 2 work, it had my average HR in the 205-210bpm range (my max HR is in the low 190s…). Saturday I was doing zone 4 work on Zwift, same watch, and I maxed out at 148bpm doing 5min intervals in Zone 4 at the end of a hard interval session (preceded by long Zone 3 efforts)…
It’s worth noting on that 205-210 day, I took my own carotid pulse to verify what the Garmin was telling me, and I was getting HR’s in the range of 120-125bpm…
. . . assumes a fixed cycling economy . . .
Thanks for the explanation. The fact that they assume a fixed economy factor would seem to make the measure less than useful. After all, one of the primary things I want to do in my training is to increase my economy.
According to FirstBeat’s whitepaper (the algorithm provider behind Garmin’s fitness/recovery tools), the error is +/- 4.3% when true HR max is entered.
https://www.firstbeat.com/...2max_11-11-20142.pdf
Importantly, it looks as though the algorithm accounts for general pace variability but not grade in its calculations.
Also worth noting that for those that have darker skin, the optical HR tends to struggle (they need to use infrared, not the green lights to improve accuracy). I have a 235, and recently put it into broadcast mode for a Zwift ride, and for Zone 2 work, it had my average HR in the 205-210bpm range (my max HR is in the low 190s…). Saturday I was doing zone 4 work on Zwift, same watch, and I maxed out at 148bpm doing 5min intervals in Zone 4 at the end of a hard interval session (preceded by long Zone 3 efforts)…
It’s worth noting on that 205-210 day, I took my own carotid pulse to verify what the Garmin was telling me, and I was getting HR’s in the range of 120-125bpm…
I doubt the ability of an optical heart rate to accurately measure HRV during vigorous exercise.
It is way off
Garmin Fenix 5, running VO2max:51(today)
Test in a lab, running on a huge threadmill :60,75 (~2 months ago)
.
Speaking of VO2max, we did do this test in the lab the other day:
As others have mentioned, Garmin generally relies on Firstbeat for performance metrics and these are often based primarily on heart rate variability (with knowledge of age, gender, mass, height, etc.). These are correlations, not direct (or indirect) measurements. So, an accurate correlation of the entire population is tricky. Ultrafit, fit, average fitness, limited exercise, couch potatoes - well, Garmin will have customers across the range. From elite endurance athletes to grandpa recovering from an injury and shuffling through the mall. Decisions are made wrt the correlations and some cohorts are better represented than others. Also understand that the correlations are based on the tested population. You would be shocked (well, ok, maybe not shocked) that many of the correlations are derived from ‘convenient’ participants. What do these participants ‘look like’? At research universities, they tend to be moderately fit 19-22 y/o (i.e. - undergraduate students). Is that you? At Garmin and Firstbeat, the participants might be older but pretty fit. Is that you? Metrics are extrapolated from these populations of relatively homogeneous characteristics to all comers. The fact that the reported values are pretty darn good for a large fraction of the users is remarkable, IMO. FWIW, my own Garmin/Firstbeat VO2max is very close to indirect calorimetry in a research lab. 57 y/o male, pretty fit but far from elite (MOP nationally, top 25% regionally, podium locally (yeah, I live in an unfit portion of the country)). 5’6", 125-130 lb, so at the small end of the metrics for height/weight (for males). For some, the Garmin/FB predictions are off. I’m surprised that they are as accurate as they are for me, someone quite distant from the characteristics of the cohort on which the correlations are based.
Your Vdot02 will be much more indicative of your actual predicted finish times.
It is also lower than V02 Max.
This is coming from a Forerunner 920xt/Edge 520 but I assume it’s all the same with regards to how it determines the Vo2max values. But for running the 920xt guessed it right at the same Vo2max as the one I got during lab testing (I make sure I keep my weight etc up to date on the Garmin app). My Edge 520 however detected a Vo2max during indoor intervaltraining on the trainer that is 8-9 units higher than what I labtested for running… So it might be coincidence that it actually got it right for running.
According to FirstBeat’s whitepaper (the algorithm provider behind Garmin’s fitness/recovery tools), the error is +/- 4.3% when true HR max is entered.
https://www.firstbeat.com/...2max_11-11-20142.pdf
Importantly, it looks as though the algorithm accounts for general pace variability but not grade in its calculations.
The fact that it doesn’t include grade in its calculations is a huge limiting factor as to how useful the V02 numbers are on a Garmin. I would also add temperature to this. Every time I go for a favourite mountain run that is 28km with 800m of ascent, I come back to “unproductive” as it lowers my VO2 estimate based on the pace. Generally, I run this 28km at about 5:00/km and if I were to go out for a flat 28km training run, I would be 4:30/km. I get a similar indicator when the temps increase, along with HR, but my pace remains flat or slower.
This makes the training status indicator wonky. I know that it’s not entirely wrong, but it’s often way out too, so I never trust it. The indicator that shows up after a km or so of running indicating baseline fitness (+ or -_) tends to be a pretty good indicator of how things will go on a particular run. But never look at this number at the start of the run in a tri! 
Your Vdot02 will be much more indicative of your actual predicted finish times.
It is also lower than V02 Max.
VDOT is essentially just a different way of expressing your finishing time(s).
Despite the somewhat misleading name (V-dot-O2-max) should not be in any way conflated with laboratory-measured VO2Max.
In simplistic terms, the former is essentially the output of the latter, with athlete-dependent fitness, running economy, biomechanics and race execution dictating the degree to which one translates to the other.
To circle back round to the OP, the Firstbeat/Garmin method tries to model VO2Max using HRV measurements and then uses a system similar to VDOT for it’s race prediction times; these times aren’t a direct copy of the original Daniels and Gilbert paper (and probably rightfully so, given that study used elite middle and long-distance runners as subjects) but evidently a similar method.