Altimeter+Bike on wrist

Possibly the difference could come how wind hits the barometer sensor. The hole for barometer sensor is on the right side of watch, so it is aligned differently in relation to your riding direction whether it is on your wrist or your handlebar.

that could explain a very small variation of a few meters, but not likely the differences the OP is describing.  FWIW, I have a fan that can blow air at about 25mph and when I put my Fenix at various angles directly in the airflow I could only get it to vary by 10 feet or less as the wind flowed over every possible angle.

There is clearly a problem in the software of how elevation change is calculated.
I did a run on a flat track today - a 300 meters track, so the track run mode didn't ackowledge the track - and activity gave me a 74 meters ascent on a 10 km run... Barely 35 laps. definitely impossible. What is interesting is that it also detected 75 meters descent. So pretty consistent in the elevation method... But software needs some more aggressive smoothing in my opinion.

I get the impression that people here expect way too much out of a wrist-based altimeter, they seem to expect lab-quality performance out of a watch, which just isn't realistic.  The 1m resolution of the altitude app display will bounce around a little while stationary, that's to be expected with an instrument of this type, and those fluctuations are reflected in the ascent/descent running totals.  The fact that over time they show essentially a 0 bias seems to indicate that the altimeter readings are pretty stable, given its extreme sensitivity.  At sea level 1m of elevation change is about 0.0017psi of pressure change, or 0.00012bar depending on your preferred units, its very small.

I never run on a flat track so I can't compare your experience directly with mine.  I run short 5k runs on a street that has about 20ft of elevation difference from one end to the other according to Google Earth.  In 8 laps I generally see ~150-160ft of ascent and descent (they're usually within a few feet of each other), so mine tracks the elevation change pretty accurately.  If I subtract the actual elevation changes I encounter during the run, the residual is only a few meters, which does seem considerably better than what you get.  Have you checked the course you run on Google earth to see if its not quite as flat as you think, if there's a meter or two difference from one end to the other that would add up over 35 laps.

If you filter it too much then it wouldn't respond quickly enough while riding a bicycle, or driving in a car, but it might be OK for walking or running.  So, there's no magic filter setting that's going to work best for all possible environments, tradeoffs have to be made.

drkglb said:

definitely impossible.

Why? Are you certain the track is dead level? Is ~2m per lap so unlikely?

7040565 said:

The same routes give 100m of altitude difference.

Over what distance and what overall elevation?

Dr Phil said:

Why? Are you certain the track is dead level? Is ~2m per lap so unlikely?

Pretty much. The track is brand new, it is built on a flat terrain so I am confident it is completely flat - even though, as you guessed, I didn't measure it myself

5693709 said:

I get the impression that people here expect way too much out of a wrist-based altimeter, they seem to expect lab-quality performance out of a watch, which just isn't realistic.  The 1m resolution of the altitude app display will bounce around a little while stationary, that's to be expected with an instrument of this type, and those fluctuations are reflected in the ascent/descent running totals.

You are right, maybe it's just a normal error range - but I remember doing lots of track runs with previous models (Fenix 5S, 5S plus) and always ending up with 5 to 10 meters ascent which FELT far more realistic. So I can't help buth think that +1 / -1 meters elevation changes over a 200, or 400 meters distance could easily be ignored and not accounted. When this value is higher (let's say at least 5 meters, which should be enough) the ascent is accounted. That's what smoothering is basically.
Otherwise, I'm overall happy with the watch and this is no big deal. On normal runs, the ascent is usually on spot with the expected values which is great..

Almost 75km.

I think you're right. When the watch detect pulse on wrist elevate the meters you up. 

I used 2 times at home in the balcony with roller and detect gps and climbed 24 meters without move...but I used on handlebar and 0 meters. 

I mailing garmin but only sayed restore the watch. Great solution...