A better way to handle bad Segment data, especially elevation data

Simplest suggestion:

  • Let users hide segments from the segment page (i.e. Actions->hide), don't make users have to go back to the activity page collapse the segment view and then click hide

Hopefully more users will then hide bad segments!

Next suggestion: Focus on getting categorised climbs correct. Incorrect cat climb segments is the worst part of all of this (IMHO) and perhaps just addressing these is sufficient;

  • Let users flag incorrect cat'd climbs! (either an option to the existing flag action, or a new action entirely)
  • Once a cat-climb segment gets 10 users on it, compare elevation profiles from each users activities (along with map data) to confirm it is a cat climb.
  • Run super clever algorithms (I know you can do it!) over all cat climb segments spot outliers (i.e. avg speed to high etc.) and evaluate those.

Most of the above could be extended to all (not just cat climb) segments, but obviously a bit more work etc.

 

Getting better at the above would be amazing and make the segment explorer so much more useful,

 

Thanks!

 

 

 

12

Comments

17 comments
  • Thanks Chris! 

    I hear what you are saying about making the "hide" process easier for segments. Thanks! 

    In regard to segment elevation, largely the problem is with auto-generated segments from data with poor barometric altimeter data. We've since disabled automatic segment creation, but there are still quite a few segments in the system with poor elevation.

    If you send us a list of segments with incorrect elevation via a support ticket, we can correct the elevation for you. 

    Hope that helps! 

    -1
    Comment actions Permalink
  • One very simple method Strava could use to find bad segments is simply searching your database for segments with rides having huge VAM numbers, like 3000 or greater.  I see these all the time and usually the segment was originally created with junk elevation data.

    1
    Comment actions Permalink
  • Is it a bit cheeky to put in a support request to correct elevation data on loads of segments at once?

    I have put in support requests in the past when I find that I've ridden a segment that is flat or downhill but is listed as Cat 4, but looking on the Segment Explorer just now, I see at least twenty categorised climbs in my relatively flat hometown alone and want to know if it's okay to list them all in one request or not. It is going to mean a lot of work for the poor Strava employee who gets that request!

    0
    Comment actions Permalink
  • I would also welcome some proactive correction of bad elevation data by strava. In the area where I live, which is relatively flat, I could easily identify 100+ segments with very bad elevation data resulting in steep climbs (>10%). Often these segments had a negative lowest elevation, so these bad segments should be easy to find for strava.

     

    I have created a support ticket for the almost 60  segments ids which I will likely encounter during my rides. Hopefully strava support will do something with the reported data.

     

    1
    Comment actions Permalink
  • Steven and Pascal - we're also working on a system to better handle the reporting of bad segment data. You might see changes as early as August! 

    0
    Comment actions Permalink
  • After my request of recalculating bad elevation data I found a side effect. 

    My biggest climb was one of those bad segments. After correction the biggest climb on the my profile page now showed the correct elevation gain for that bad segment. Not a big deal since I expected that after my next real climb that segment would be wiped out. Surprise .... it was not. It seems the incorrect elevation data is also stored in a separate table which is not corrected.

    This is something that should be taken into account when correcting bad segment data in the future.

    1
    Comment actions Permalink
  • We've released some changes to the Segment Flag system on Strava.

    Please check on a segment currently flagged as Hazardous to see the new options available. You'll also see changes to the flow for creating a new hazard segment flag, which allow users to report issues other than a segment hazard. 

    I'd appreciate all feedback about the hazard segment flag updates to be posted to this product feedback forum:

    https://strava.zendesk.com/entries/47450040

    Thanks, and I look forward to your comments! 

    0
    Comment actions Permalink
  • Hi Elle,Those improvements look good and should make handling hazardous segments much better.

    However, as you yourself note in the comments section for the feature, they don't address the issue of bad elevation data - but they come sooooo close.

    If I understand correctly, if I flag a segment with the reason 'Bad elevation data' then Strava just ignores that entirely, is that correct? The message it comes back with suggests that:

    "You have chosen to flag this segment for: "Bad GPS, Elevation or Duplicate". Unfortuately [sic], segment flags are reserved for hazardous segments only."

    It feels like it wouldn't take much for strava to record bad elevation flags and take action based on them (as suggested by Steven in the feature comments), instead of requiring users to create a separate support case.

    It would be so great if Strava could sort this out, it's really disappointing that I can't use the segment explorer tool to find good climbs in the area due to all the false positives!

    Thanks

    0
    Comment actions Permalink
  • Agreed Chris! We are so close to having a tool to address bad segment data. And you are correct, we don't allow segments to be flagged simply for bad data at this time. The current workaround is to contact our support team with the link to the segment, and we'll do our best to fix it. 

    0
    Comment actions Permalink
  • Just a Tip. The GPX uploaded file, does it have the GPS system in use by the user? If it is recorded by an outdated GPS.. or maybe if is recorded with a GPS with a barometric altimeter? If is so, you eaily can overlap the elevation data fixed soon as user with Barometric Altimeter uploads data to strava... 

    0
    Comment actions Permalink
  • @Jose - Indeed that is true. We have a tool available to the support team that can take the barometric elevation data from another file, and add it to the segment profile. 

    0
    Comment actions Permalink
  • I would estimate that in my area there are 90-95% false positives for cat. 4 climbs. To find true cat 4 climbs with Strava I would have to spend more time on Strava than on the bike. This is very frustrating, and one of the main reasons I didn't keep my premium membership after the trial period ended.

    0
    Comment actions Permalink
  • An observation/suggestion:

    The majority of falsely categorized climbs have "climb" in the title, and the elevation data starts at a negative number... It seems to be always somewhere around -200ft.

    Many of these have only 1 (or very few) attempts made.

    It seems to me it would be easy to filter these out.

    At least give the end user a way to hide these on the Segment Explore maps.

    1
    Comment actions Permalink
  • I agree with Mike Wilkens that the "col category" filter in segment search is useless now. (I also see that some segment creators seem to be aware of the data being bad.)

    Solution could be:

    • Users can report segments for bad elevation data
    • Trigger check if 10+ reports (or whatever number)
    • Trigger check if VAM > 2000 (or whatever number)
    • Check by a human
    • Check by computer based on map elevation data
    • Check by computer based on GPX data from other riders

    Or any combination of the above.

    1
    Comment actions Permalink
  • It would be nice for users to be able to hide or filter climbs and that's quite easy to implement within the next update.

    https://dissertationwriter.org/

    0
    Comment actions Permalink
  • Once again, asking for a way to flag segments with erroneous elevation data. It is irritating to see that a "400ft climb" is my biggest climb ever when, in reality, it's maybe 60ft tops.

    1
    Comment actions Permalink
  • Strava has the computing power and I expect the historic ride/activity data to be able to make sense of the elevation data on segments and they would certainly have someone who understands statistics. Even if they can't publish the "actual" elevation down to the nearest 10cm surely they have a statistician smart enough to apply an algorithm to their big data engine to work out that a single highest elevation measurement of 6000 m should be ignored on a segment were 98% of elevation data for that segment says the highest elevation is 89m. But they don't seem to want to do it for some/$ reason. Statistically the lowest/highest elevation (and gradient) should be the median value rather than the mean/average value to remove distortions from those outlier values. The algorithm could/should be applied every week/month to any segment where there are less than 1000 data points and once a year to segments with over 1000 data points. 

    1
    Comment actions Permalink

Please sign in to leave a comment.

Didn't find what you were looking for?

New post