Garmin Connect Courses elevation gain grossly overestimated.

This issue exists for years. Workaround: use Strava with its automatic synchronization that pushes a route to your watch or create the track somewhere else and copy the GPX file directly to the watch (with cable)

I think that we need a solution instead of a workaround: Garmin should solve it and give us a real working software! Come on Garmin, you can do better than this!!

I can also reproduce this. Find below the link to a gpu file. From my point of view these are two issues:
1. The Garmin course tool is off by 20 - 30% when calculating ascent/descent

Here Garmin Connect: 1869m of ascent

Same GPX imported to the Swiss Topo Map gives 1181m of ascent - that is pretty much what I also get after riding.

2. When syncing the course to the device, the ascent/descent values get doubled (you can only reproduce that by adding a course to your garmin connect and sync it over. When I manually convert the GPX with GPSBabel and copy the file to device on my Mac, everything is fine. To reproduce, import the gpx file into garmin connect and then sync it to the device. 

GPX-File: https://www.dropbox.com/s/xafrsry5di1v4sq/Hirzel%20-%20Horgen%20-%20Sattelegg%20Loop.gpx?dl=1

FIT-File converted with GPSBabel: https://www.dropbox.com/s/h6guj334urnojwq/Hirzel-Horgen-Sattelegg%20Loop.fit?dl=1 

I opened a ticket with garmin support and never heard a single word back about this for about half a year... still, I would encourage you to do the same. 

I can also confirm this has existed at least a few years. Courses created in Garmin Connect largely over-reports the elevation gain. This is compared to most other route creators (ridewithgps.com, strava.com, etc), and, importantly, compared to the elevation gain actually measured by Garmin devices with a barometer after having followed the course. It's rather mindboggling that Garmin, originally known for their navigation tools, are screwing this up so badly compared to everyone else.

torsoli82 said:

It's rather mindboggling that Garmin, originally known for their navigation tools, are screwing this up so badly compared to everyone else.

The true question is who is right. On all the courses I verified, the calculation was accurate. The problem is rather that Garmin is more accurate than others, and more accurate than the barometric altimeter, since it uses a quite high density of track points, and takes in account every small difference between neighbouring points. So even if you ride on a flat terrain, it adds all the small bumps to the elevation gain, although the barometric altimeter won't notice them. And other tools either use lower number of points, or use algorithms smooting out the bumps. Whether it is righ or not, that's disuptable. The best would certainly be, having an option for it.

It's been quite a while, when I report this to Garmin. At that time they told me, that they are aware of this issue ad that they are working on it ...

its been almost10 months now and i never heard back on that opened ticket.

My experience tells me that there possibly are geographical differences between how accurate / inaccurate Garmin Connect's elevation gain is. I don't know which courses you have verified, and how exactly you have verified them, but I am fairly confident Garmin often is way, way off in their reported elevation gain.

Mario Schaniel have provided one good example just above, using both his barometric recording and SwissTopo, Switzerland's national mapping agency, whose Swiss maps are praised for their accuracy and quality. 

But let me give another example from outside Switzerland. Look at this ride in France, starting in the coastal town of Menton, up and over Col de la Madone, Col de Braus and Col de Turini, before heading back to Menton: https://connect.garmin.com/modern/activity/3902686621. It was measured as 117 km with 2946 m elevation gain by my Garmin Edge 810 bike computer. For information, the bike computer of my riding partner that day recorded an elevation gain of 2937 m, i.e. almost exactly the same. 

Now compare these data to the course I drew in Garmin Connect a few days before that ride: https://connect.garmin.com/modern/course/26478986. Garmin Connect reports 115 km with 7258 m elevation gain. 

2946 m vs. 7258 m. Do you still stand by your claim that Garmin is so much more accurate than all the others? Do you think this route includes 7258 m of climbing?

Further, as an experiment, I just redraw the very same course in Garmin Connect now, about 1 1/2 year after I drew the first one. Here is it: https://connect.garmin.com/modern/course/48306631. Interestingly, it now it reports 6734 m elevation gain. I.e., for reasons unknown to us, Garmin has reduced the elevation gain of this course by about 500 m over the last 1 1/2 year. So which one is the "very accurate" one, the one that is so much more accurate than the all other sources? 

For added information, if you draw this course on other platforms, you get something that is not identical to, but much closer to what is measured by the barometers, e.g. 3492 m here by ridewithgps.com: https://ridewithgps.com/routes/34972250.

I've ridden this loop, and my personal anecdotal feeling is that there is no way you're climbing ~7000 m during it.

torsoli82 said:

I don't know which courses you have verified, and how exactly you have verified them, but I am fairly confident Garmin often is way, way off in their reported elevation gain.

I checked some of the Courses posted in this thread. The method was verifiyng the elevation of individiual keypoints, converting the course to a CSV file, importing it to Excel, and summing up the elevation difference between neigbouring points. 

torsoli82 said:

2946 m vs. 7258 m. Do you still stand by your claim that Garmin is so much more accurate than all the others?

Yes. It is more accurate in the sense that it includes even the small differences between individual keypoints. If I tell more accurate, it does not mean it is the better solution, because the differences between individual keypoints may be 1) just small bumps, that other applications (and barometric altimeters) do not register, and 2) may come from certain accuracy tolerance / fluctuation of the topographic data. So yes, I stand by the claim it is more accurate in that sense that it does not filter out these small differences, while most other appplications filter them out (or use topographic data of lower density, greatly reducing the problem too).