Mountain Bike Reviews Forum banner
1 - 20 of 24 Posts

·
Registered
Joined
·
609 Posts
Discussion Starter · #1 ·
GRR! So I just found a really frustrating bug with Strava. I ride Waterdog at least once a week, and I've been crushing myself to move up the latter on various KOM segments. But what I just realized, is that if the trails are close enough together, it will categorize part of your ride as an existing segment, even though it doesn't match the route you rode. For example, there is a route labeled DH John Brooks-Chaparral-Rambler. But when I ride John Brooks all the way to Rambler and down to the parking lot, it lumps me in to the segment that cuts off of John Brooks to Chaparral and over to Rambler.

So, I created a new segment. I thought, problem solved, no big deal. But now the KOM leader for the segment I created has a significantly faster time. Well, he took a different route down to the parking lot, not the route I specified in the segment I created.

I don't know, part me thinks I shouldn't care. But I got really excited about Strava, paid the $60 membership, and now I feel like the excitement of the idea of King of the Mountain is sort of gone. 90% of my riding is solo night riding at waterdog since it's what my schedule permits, and I really got excited about turning a solo ride in to virtual XC race.

Has anyone else run in to this issue? Waterdog is a pretty small park with trails that parallel each other close together, so maybe it isn't an issue elsewhere.
 

·
Registered
Joined
·
121 Posts
I can see how that would happen. You should report the specific cases to Strava and they can look into manually fixing them or improving their algorithms. The problem is that GPS data is not precisely accurate. They have to allow for some error margin when matching rides to segments, otherwise they would never match up. I can imagine that they match the start and end points of the segment to the riders' activities, and probably some points in between. As you suggest a particularly difficult case would be two segments that start and end in the same places and have slightly different routes (eg. 2 parallel tracks) - if they are run close enough then it would be very hard to disambiguate them. This is much less of an issue on the road, since roads are much wider than trails and roads don't run as closely to each other. Also most segments may well be longer than your MTB segment.
 

·
Registered
Joined
·
609 Posts
Discussion Starter · #4 ·
Yeah, I posted something to their support blog. I guess part of my frustration is that I do understand the difficulty with what they are trying to do on short segments where multiple trails intersect each other. But if it doesn't work, it drastically reduces my interest in Strava and I would definitely not sign up again. Hopefully they can fine tune it a little.
 

·
I like mtn biking, too
Joined
·
3,259 Posts
I have not experienced that problem, but I do notice that when I do a couple loops on the same (narrow) trail, my tracks on the map from each loop did not overlay one another exactly. (See the southern end of Boggs where it's not even close --> http://app.strava.com/rides/303941) In that particular trail system, the tracks were not too close to other trails or segments close by, so it wasn't really a problem. I can see how it would be in a place like Waterdog, though. The GPS is what is reporting it that way. Strava just is showing what the GPS says.
 

·
Registered
Joined
·
3,486 Posts
They are pretty responsive if you flag a segment. I had them change segments that were not correctly labeled. I'm not familiar with Waterdog, so don't know if it would apply.
 

·
Registered
Joined
·
609 Posts
Discussion Starter · #9 ·
So I got a reply from Strava. It doesn't look like it matters at all how close the trails are to each other. They only consider the start and end points.

"Hi Matt,

Yes, you're right, seems to be a false match. For parallel segments to routes, its difficult because we don't currently use intermediary points to match a segment. Just the start and end points. So, this sort of thing happens from time to time.

We're aware of this, and its on our list of things to fix and improve. I've removed the segment from your particular ride, and we'll have to wait until the issue is resolved to prevent this from happening all together.

Thanks,
-Elle"
 

·
Registered
Joined
·
3,486 Posts
maleonardphi said:
So I got a reply from Strava. It doesn't look like it matters at all how close the trails are to each other. They only consider the start and end points.

"Hi Matt,

Yes, you're right, seems to be a false match. For parallel segments to routes, its difficult because we don't currently use intermediary points to match a segment. Just the start and end points. So, this sort of thing happens from time to time.

We're aware of this, and its on our list of things to fix and improve. I've removed the segment from your particular ride, and we'll have to wait until the issue is resolved to prevent this from happening all together.

Thanks,
-Elle"
Most of the segments are very short. Fastest point to point makes sense.
 

·
Registered
Joined
·
609 Posts
Discussion Starter · #11 ·
EBrider said:
Most of the segments are very short. Fastest point to point makes sense.
Not really. The short ones seem to be the most accurate. It's the longer routes with multiple options that get messed up. The route I'm most interested in is ~20 minutes. Fastest point to point would be to bomb down fireroad, then take pavement to the main lot bypassing all of the climbing and fun singletrack. Oh well, it's not the end of the world. Hopefully they'll be able to get it resolved at some point. Either way, it has still added a level of fun and excitement to my rides.
 

·
Registered
Joined
·
744 Posts
They will fix it, just give them some time. Strava has been great to me and all the issues I have had or features I have asked for over the past few years. Let them know the issues and they will work on fixing them. They are by far the best place for ride uploads in my opinion and if you knew where they were a year a go you would trust that they are working very, very hard to make it a great place to save your rides.
 

·
Registered
Joined
·
1,751 Posts
I've noticed the same problem in Fremont Older.
My recommendation is you take the same short-cut as the KOM leader to try to get the title back. Otherwise, you can always just go for your own PRs on your original route.
 

·
Registered
Joined
·
5 Posts
Great Discussion!

First off - I understand the frustration of having your effort compared to others that took a slightly different route - and this is exactly the type of thing we'll keep working on. It may never be perfect, but we'll keep improving the segment matching.

Here's a quick overview of how segment matching works. Of course the start and end must match, but there are also several points along the way that must match as well - based on an algorithm of ours which reduces a segment to a number of critical and required points along the way.

There is a tension and balance between false negatives and false positives when it comes to segment matching. The false negatives happen when a segment isn't matched to a ride when it should. This usually happens when the matching requirements are too tight, and / or the GPS had a drift at a critical point along the segment. False positives are when a segment matches when it shouldn't - as this thread is about. These happen when there are trail or road systems that are very tightly consolidated, and the matching can jump from one trail to another when it's very close in proximity.

It's a tough one to solve because if we tighten the matching requirements to decrease the false positives - we tend to get more false negatives (segments that should match but don't) and visa versa.

What seems clear is that some areas - like Waterdog, may require a segment matching engine that is tailored to it - and there may be some ways we can use crowd sourcing to auto improve such an engine by enabling users to point out both false positives and false negatives for a particular segment.

What I can say for sure is that we have some very smart people thinking hard about this issue, so, stay tuned.

Again - sorry for the frustration, but all solutions originally are inspired by a problem.

Davis Kitchel
Strava.com
 

·
M8 M12 M15 deez nuts
Joined
·
8,610 Posts
The only frustration I have is in the form of confusion: there is a guy who shows up on the MTBR team activity page by the name of Ron Shevock, however, none of his footies are being credited to our team since he is ACTUALLY not on our team, but instead on the Team Madcat team. This seems to be some sort of Strava bug.
 

·
Registered
Joined
·
609 Posts
Discussion Starter · #18 ·
Leopold Porkstacker said:
The only frustration I have is in the form of confusion: there is a guy who shows up on the MTBR team activity page by the name of Ron Shevock, however, none of his footies are being credited to our team since he is ACTUALLY not on our team, but instead on the Team Madcat team. This seems to be some sort of Strava bug.
LP, you are confusing Strava clubs with the team you participate in the KOM challenge with. For example, I am member of both the MTBR and Team Wrong Way "clubs", but I log my feet for the KOM challenge with Team Wrong Way.
 

·
site *****
Joined
·
101 Posts
Davis, thanks for the reply. The website is very impressive!

Ron Shev (Tyrone) is doing the KOM footies under the Mad Cat club affilation. I tried to talk him into switching today.

My bug is I am not listed in the Overall Men KOM list.
 

·
M8 M12 M15 deez nuts
Joined
·
8,610 Posts
OK, I guess I am confused on the whole affiliation thingie, I had assumed one pledged allegiances to one team entity. It is confusing seeing what he is doing, yet not seeing his footies being credited to our KOM effort.
 
1 - 20 of 24 Posts
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top