Just analyzed 2 years of run and seems messy to me :-/. Is there any way to remove the routes?
Also I have runs when I only cross the oval, but sometimes I run one or more laps on the way and they are all recognized as same route. But that is understandable.
This is worse:
Can you please msg me a link to that one. Tx.
Also got one idea if it´s aplicable. I´m not good programmer⌠There could be another condition to compare route vs. activity track. If You can take (from GPS points coordinates) the most âtopâ, âleftâ, ârightâ and âbottomâ points, You can get the rectangle. The same for route. And then the rectangles edges shouldn´t be far more than 20 meters e.g. That will eliminate bugs like this.
Great feature and I analysed a bunch of activities: it seems to work quite well.
One thing I think about is how to handle races with about the same course each year. Should the name of the route contain the year of should there be a generic name for the route and then add the year? Some races alter their courses slightly each year and then itâs a blurry line when to consider one route different from another.
Tx. Using a generic name for the race and manually adding the year should work well.
alright - guess I just ran into a server-side timeout or something ⌠maybe i was a bit to enthusiastic in wanting to test it out ⌠
ps) can you add Route / Route ID to Filters? So we can Filter already processed ones? 
ok - no timeout - found some activites which crash the route analyze:
â NullPointerException
ID i39779740
ID i41946447
ID i43211510
ID i49271687
also got a little Feature Request regarding this cool Feature - maybe in the Future when you find the time something like this could be implemented!
It would be fantastic to have a âTop Routesâ overview page that displays the routes Iâve completed most frequently. This could include details like:
Basically, a detailed breakdown of all available data for each route. I think this would offer great insights and be really exciting to explore!
Thanks for considering this idea! 
I have some activities that are the same, but the routes donât match, because sometimes it takes longer to acquire the GPS, so the distance varies.
and
Oh, and because of road works/my desire to end the activity outside a shop and buy a recovery drink, the end also varies slightly. Could there be an option for route tolerance? In know strava also refuse to manually match segments if they differ too much.
I am going to add that. However the âOnly do routesâ analyse option doesnât do anything to activities that already have a route so you donât have to filter them out. It just slows things down a little.
I have just fixed that. Tx.
Yes that would be nice and a natural progression of this feature.
This one is tricky to implement. The database index for the routes includes the end point (likely GPS is working by then) and distance and canât be too fuzzy or too many routes have to be checked for a match for each activity.
The other problem is that when no route is found a new route is automatically created. And these are shared between all athletes. So if someone in your area had their route matching to be less fuzzy you would end up matching their routes even if your route matching was set to more fuzzy.
sounds great!
Guess those 2 should match the same Route - but arenât?
i39777551
i39777543
Nice feature, looking through the routes, one I noticed is that 11057 and 11053 should match, if Iâm not mistaken. Same for 11149 and 11170.
Yeah, but this is the thing with algorithms: a human can easily classify these routes as the same/quasi same.
Hi David, still having a few issuesâŚ
These activities have been matched as the same route but arenât.
One is 8.49km and the other 8.13km. That is (just) different enough to not match.
[quote=âalexpeits, post:32, topic:86792â]
I noticed is that 11057 and 11053 should match[/quote]
The finish points are a bit far apart but is a very close thing. I uploaded a run from each of those routes to my laptop and they matched the same route. Different import order tipped it.
Finish points too far apart.
Thats a tricky case. The shorter one got done first and created the route. The longer one matches the route very well with a bit extra tacked on. Intervals.icu relies on the distance check (within 4% or 200m whichever is bigger) to catch that. 4% of 12.82km is 512m so the longer one fits (320m different).
If I make the distance check tighter then lots of routes that look like they should match, will not. I started with 2% and bumped it up after testing on my own (cycling) and my wifeâs (running and a bit of cycling) data.
Nice because it outputs a similarity score.
Iâve got a daily commute that also includes a train ride, so probably too tricky a situation. I also make it one activity by pausing it during the day so the start and stop times are all over the place. This is probably not going to work as a commute route. I did run the analysis and I got the following Route IDs:
12551, 12568, 12569, 12570, 12638, 12639, 12640, 12641, 12642. These are all basically the same commute for me.
Yes those donât work too well. However you can still tick the commute flag for each route and have the activities automatically flagged as commute.
I added a boundary box check to catch this case. If the ratio of the activity width to the route width is more than 1.1 then it isnât considered a match. Same for height and diagonal. So of you change a route inside by taking a few different paths it will probably be considered a match but if you add on extra on the edges then it might not.
I had a look at some of yours and you donât end up with too many routes. The âlabelling commutesâ use-case will still work, you just have to label more routes.
