Garmin Fenix 6 Pro distance measured much shorter than Vivoactive 4 and Google Maps

How do you wear the watch? Like normal on the outside of the wrist? Then you can always try to wear it on the inside wrist to see if it makes any improvements. For me I'm usually getting about half the distance loss if I wear the watch on inside wrist instead of on the outside wrist.

Also, if you export the activity .GPX file from Garmin Connect Web and then import the .GPX file again. Do you get a longer distance then?

I know this was not addressed to me but I have never been able to reproduce the short distance issue, going right back to when I posted Polar V800 vs F6 Pro results in the long running thread to which you and I have contributed.

From today's run I just got back from (I am actually testing wrist HR for this run not distance btw)

I really think at this point, it's more than reasonable to assume people who have shortness issues or runs recording too long, way over and above the GPS error margins have faulty devices. 

I wear it normally, outside of the wrist. I'll try what you mentioned but I think it will not be convenient...

I exported yesterday's activity in GPX and imported that back... You are right, distance is longer, closer to the real distance covered. How do you explain that?

The distance from the watch is not only from the GPS. Its also about running technique, varied running speed, varied running force etc etc. The watch seems to have some "smart" functions in both the distance and pace algorithms which gives wrong numbers for some people.

You can always try to wear it on the inside wrist once to see if it makes any difference for you.

About the distance deviation between the GPX and the activity. See my answer to 5135968.

AndersB said:

The watch seems to have some "smart" functions in both the distance and pace algorithms which gives wrong numbers for some people.

Well the evidence presented has never been very compelling and extraordinary claims require extra-ordinary evidence. 

Why do you not think the above is more likely than it being a faulty device issue? - in the long running thread, and the link above, people who have had device swaps have reported the issue solved.

But 

AndersB said:

It might be some hardware failure but I personally thinks that its a software algorithm issue.

But Garmin very obviously uses a common code base - so I think an algorithm issue would manifest over multiple watches ie Forerunner, Vivo active etc. I have seen people say with absolutely no evidence whatsoever that the Fenix line is optimised for walking (!!) and the Forerunners for running. Then we also had the dis-information being spread that Garmin exported .GPX not .FIT to Strava so the distance 'issue' was being hidden from users -. I debunked that one in the long thread by showing how Strava did indeed use .FIT

If I was to speculate, and I prefer not to, I would suspect a faulty GPS antennae or faulty accelerometer is more likely than a software issue, although I would assume we'd only see accelerometer input in unfavourable GPS conditions.

Have you compiled your data field for the 255 - I will try it out if you want?

smart reading takes reading when developer decides expose new value

every second system asks 'getValue' and you can return 'have nothing' even change is quite big and no saving (saving of very expensive operation)

Yes, there are a lot of different theories about this issue. But its hard to give hard evidence about it. I have gathered a lot of data and I can see patterns in the deviations. The deviations are smaller (for me) when I walk or run the bicycle compared to running the same courses. Also I have noticed that the deviations is getting smaller (better) when I'm in good shape and runs consistent with same speed/pace, cadence and arm force. The problems are especially big when running varied in the activities.

Unfortunately my production release of the data field is rather old and do not support the 255. I'm in the middle of a complete rewrite of my algorithm and haven't had so much time for it lately. As soon as I'm happy with the new version it will be released and will then support all new models. But the new algorithm is rather complex to handle different situations so there is very much testing and tweaking from my side before a release.