Enhanced segment matching model for long segments would probably increase the fun for many
Hi Strava developers,
The reason for this suggestion is my ticket "#1268778 Segment with zero results" which answers my question why my segment over a complete race/tour doesn't match any cyclists, not even myself. The reason is that my segment is too long and cyclist taking a slight different way do not match the segment. You describe this quite well on https://support.strava.com/hc/en-us/articles/216918227-Optimizing-Segment-Creation-how-to-create-good-segments#long.
I do understand why no match was found (except for myself), but do not understand why you modelled the matching algorithm this way. Speaking for myself, but probably also for many others (up to you to confirm this) I don't care for a slight deviation from the segment in order to match it, as long as the main way matches. I could imagine for long segments, that:
- A cyclist leaves the main race/tour course to eat something in the next street or visit a bicycle mechanic.
- Rides back for a few 100 meters to help a friend.
- Makes a mistake in the route and returns after a short ride in the wrong direction.
- Rides a parallel street for a few km.
I think that people don't want to have such matching results for segments of average length, but I think that nobody cares about this for long segments. On the other hand, long segments are attractive, if they would work, for special occations like tour events, races or well-known/famous tracks.
My request: please adjust the model for segment matching so that it stays the same for short or average length segments, but allows a larger deviation when the segment gets longer. Of course I understand that this requires analysis, tests and thinking. I do not know your current model, but I can imagine that approaches could be widening the tolerance with increasing segment length or a model allowing a ride to be N% completely off the track, where N is a low number, where N=0% for short segments and e.g. 2% for long.
I would be pleased to test it on a few rides. My data: Jan Klijn, email@example.com
Thank you in advance.