Fenix 6/7 BUG: nearly 100% failure rate when tracking Strava Live segments with sharp turns

Here is an issue has been bothering for me for the last 3 years. Fenix has a really cool feature called Strava Live Segments, but it basically doesn't work for me, at least it doesn't work in cases where it would be really valuable. If you not familiar with Strava Live segments, you can star a number of segments on Strava, then those segments get synced to Garmin watch, then during a run or ride a segment pops up on the watch as you approach it. You can see it on the map and track the live progress through the segment and race with your own previous PR on that segment or with the segment overall record. That sounds great in theory!

I really like longer segments when trail running. Those require a good effort and can be considered in virtual race. Here is one one segment that I targeted in my today's run: https://www.strava.com/segments/1980977

The problem is - about half mile into the segment the watch had stopped tracking the segment declaring that I was off segment while I was right on the trail. Last year I tried tracking the same segment with Fenix 6X and the result was the same, although 6X managed to track this segment for about 5 miles before dropping off it. Two weeks ago I was tracking another long mountainous segment with my 7X and it similarly dropped from it. I can bring a lot of other examples. In general when it comes to a longer segment, the failure rate is nearly 100%. Every single time when the watch drops off a segment claiming that I am off segment, that is not true and I am in fact right on top of the segment.

I've noticed a pattern that it is extremely likely to drop off a segment in a very sharp turn like a switchback. Clearly there is something wrong in the algorithm that makes it error prone in a sharp switchback. The same applies to following a course. I've noticed that similarly the watch is likely to go off course in a sharp turn, although later it gets back on course. For segments, unfortunately, that isn't same. Once live segment tracking is stopped it is gone.

Back to today's run, here is where my Fenix 7X stopped tracking today's segment - the exact point is shown with the black dot.

Here is a Strava screenshot of the segment map, showing the same place:

Yes, the segment quality isn't great. But Strava itself has zero issues matching this segment reliably every single time. Segments are created from activities and unfortunately many devices are not super accurate. But I think the issue has less to do with the segment quality and more to do with the 180 degree turn.

Here is a zoomed out view of this part of the segment - as you can see the watch had live tracked this segment until that sharp turn. Furthermore, as I mentioned Fenix 6X managed to live track this segment much further, so the segment quality is unlikely to be an issue.

Here is my theory based on a number of observations. First, the segment FIT file downloaded to the device seems to have tracking based on a fewer points that a bit further apart, perhaps every few seconds. I think that the segment tracking algorithm tracks the distance remaining to each next segment point. As long as the distance within the certain radius and is decreasing then everything is considered to be OK. Once the distance to the next point starts to increase, the tracking algorithm declares to be off segment or off course. However, in a sharp corner it is possible for the next tracking point to end up over the corner, which might be a little behind compared to the direction of movement, so the distance to it may start increasing. Ironically, since Fenix 7 is so much more accurate, its tolerance to bad segments that cut corners, like in the case above, is even worse than with Fenix 6.

Garmin should consider increasing the tolerance and avoid being so eager to stop segment tracking or going off course so quickly. As I mentioned above, there are a lot of really great segments and we can't control what quality of data they were originally created with, especially when we talk about the mountainous terrain or deep woods. And there is no way for us to fix the original segments. Garmin has to deal with that more reliably. 

  • I think the course tracking has exactly the same issue. The only difference is that the watch can resume tracking a course after it went off-course, but in the case of a live segment, once it is gone it is gone.

    Despite huge Fenix 7 improvements in accuracy over Fenix 6, both course navigation and live segment tracking remains unreliable in a similar way, and I suspect that has nothing to do with the accuracy. During my yesterday's 18 mile I run I used course navigation the entire time, and I can't count how many times it went off course and then back on course - probably at least 50 times. In the most of the cases the off course alerts were when I either made a sharp turn and when I stopped for whatever reason. Almost every time the watch went back on course automatically very shortly after going off course, often within 1-2 seconds. That tells me that perhaps there is some problem with the algorithm - it shouldn't be thrown off by one spurious position inaccuracy or anything like that. If it is able to go back on course within a few seconds, that means that I haven't really gone off course to start with so perhaps it shouldn't fire an alert.  Having so many erroneous off-course alerts is distracting. I am afraid that I'd stop paying attention to them and then when I really go off-course I would be less alert and might ignore that. Furthermore, every time the watch goes off-course it resets PacePro. So basically PacePro is unusable due to the poor course tracking.

  • I'll keep adding evidence to this thread in case Garmin wants to look at this issue.

    Here is the latest segment tracking failure - yesterday I attempted to live track this 5.43 mile segment. My Fenix 7X started to track to successfully but then bailed after only 0.5 mile.

    This map screenshot shows my actual track (blue) vs the segment track (red):

    The watch went off-segment somewhere around these two sharp turns in the center. My actual track was never further than a few meters away from the segment track. Can anyone tell me exactly why it decided I was off segment at that point? 

  • Another run on mountain terrain - another segment tracking failure.

    As usual the watch went off segment in a sharp turn - see the green arrow in the picture below. The recorded activity track is blue. The segment track is red.

    And again, it isn't the distance from the segment that is the problem. You can see that I was further from the segment in the lower left corner but it was tracking fine. Then in the sharp turning it went off segment. It is interesting to mention that on the way back I was following a course and the watch went off course in exactly the same sharp corner. So whatever the issue is, it is clearly the same between segment tracking and course tracking.

  • Today I tested the theory that this segment (or course) tracking low reliability is caused by a deficient algorithm.

    I made a test private segment which is very simple. Basically it is a flat completely straight line for about 0.2 miles (300 meters). Then there is a 180 degree turn and another 0.2 miles back to the start.

    I synced this segment with my watch and went for a test run. The segment tracking started fine. Then about 2-3 meters (7-10 feet) before the turnaround point the watch went off segment. So as I expected I reproduced this issue on the first try!

    I am not sure what the tracking algorithm does but based on this test and multiple prior observations it is clearly confused by a sharp turn. Perhaps the algorithm does some sort of look ahead. Then when the direction of the segment changes sharply perhaps the algorithm decides that I am running away from the segment.

    I was right on top of the segment the entire time - no further than 1-2 meters (3-7 feet) from it at any point. Both the run recording and the original recording that I used to create the segment are accurate. Both are created with Fenix 7X. Also I knew exactly where the turnaround point was to make the test repeatable.

    Sharp turns aren't common when running or cycling on roads so perhaps not many people run into this issue. But sharp turns are far more common when hiking or trail running and going through switchbacks.

  • I just did an interesting test as well. Went for a small walk. Just 25 minutes long. Previously with 9.33 I saw a drain of 10% per 1hr running. On this 25 minute walk I only saw a drain of 2%. I don't have segments or strava, or any IQ thingies downloaded. So either Garmin changed something somewhere (my guess is with the connect app as it seems the only thing they can change without any downloads from me, or walking isn't as battery intense as running. I mean, I have the feeling they changed something as my watch suddenly remains on 99-98% charge for days now while I saw some drain without activities previously. Lets see if I'm able to run tomorrow and see what drainage is doing then.

  • I wanted to post an update that I can now 100% reproduce this issue. 

    My test repro setup includes a Strava segment that I mentioned above that makes a 180 degree turn. While segments with 180 degree turns aren't common (although I know at least two such segments), what is far more common is to have nearly 180 degree turns on trails with switchbacks. And the segment tracking fails on those as easily as in my repro case.

    In all 100% of cases that I tried to follow this segment, my Fenix 7X watch went off segment shortly before the 180 degree turn.

    The distance before the turn varies from a few meters to about 10-15 meters (e.g. anywhere from 10 to 50 feet).

    Here is the screenshot of the turnaround part of the Strava segment. The green arrow shows where the watch stopped tracking the segment. That is approximately 10-15 meters (30-50 feet) before the turnaround.

    Screenshot of the watch screen showing where the watch stopped tracking the segment. This is oriented with the direction of travel so rotated almost 180 degrees compared to the screenshot above. 

    This makes it clear to me that this is an issue with the segment tracking algorithm. It seems the same algorithm is used when following a course because I've seen the watch going off course in a sharp turn many times.

  • Updated October 7th, 2022:

    We are looking into this issue further.

    For anyone that is experiencing this issue using Strava Live Segments, please highlight my name and send me a Private Message. Thank you!