Minimum Segment length ?

[Ian]

I noticed that someone recently asked that the minimum length of mountain bike Segments be decreased and the post was quickly archived by an admin on the basis that reverting Strava changes could not be discussed.

I have some sympathy for the poster - I was also somewhat frustrated by the recent increase in minimum length of bike Segments from 300m to 500m.  In my area there has been some good competition over Segments about 400 to 450m in length but such Segments can no longer be created or edited (the change happened as I was working on a new Segment of about 450m - it now finishes closer to a T junction than I would like but editing is no longer possible).

I don't remember seeing any requests from Strava users for such a change in the previous Strava Support forums so I wonder if someone at Strava could give some background to their reasons for making the change and their methodologies for determining minimum Segment lengths for each Activity type ?  Clearly measurement accuracy is a factor on shorter Segments, but even on a 300m bike Segment the accuracy should be within a couple of % which seems OK.

[eric]

I'd be interested in hearing strava engineers perspective also. Similar to your guess, my guess would also be that its an accuracy issue. There are some pretty big differences between various GPS tracking devices, and I suspect the margin for SOME users is much worse than a couple of % when looking at such a short segment. Say the fastest time is 30 seconds on a 300m segment (22.36MPH). Depending on a GPS ping here and a GPS ping there... I think some riders will see 5-10 seconds of variation which would render the segment pretty difficult to fairly measure. That's probably an exaggerated example, but that's my guess. My other guess is that they had to draw to line somewhere to prevent the segment database from getting overrun with tiny segments, or drowning out more "important" segments. 

[Ian]

As I'm sure you are aware, Strava bike Segments are numbered sequentially in the order they were created.  I started using Strava in 2017 and the first Segment I created had a number around 17,000,000.  At that time it was clear (from trial and error) that the minimum Segment length was 300m.  The latest Segments are numbered around 32,500,000.  I have completed quite a few Segments that are shorter than 300m and the highest Segment number of these is around 14,000,000.  Hence I infer that Strava set a minimum length of 300m in about 2016.  This seems perfectly reasonable - I have completed Segments that pre-date this change that are as short as 45m (the KOM is 6 seconds !) and these are clearly not suitable for meaningful competition.  

I agree that a minimum Segment length is appropriate - I'm just curious that after it being 300m for about 6 years why has it been decided to raise the limit to 500m ?  I've heard that some GPS systems only record every 5 seconds but, despite looking at many Strava Activities, I don't think I've seen an Activity with such large step intervals (by counting the data points in my Activities it seems my own GPS interval is about 1 second, despite it being from a very old phone).

The number of Strava users and Segments is no doubt increasing and so there is ever more data to be handled by Strava - however, the cost of data analysis and storage continues to fall so I doubt this would be a reason for the change. 

[Silentvoyager]

Strava could improve accuracy of segment timing (quite significantly) without raising the minimum segment length by implementing interpolation of segment timestamps. Even when someone uses smart recording and has recorded points further apart it is hard to imagine that person changing the speed drastically within a few seconds, so interpolation could help to determine more precisely when they went by the segment start or finish. Also, a smaller radius could be used for the interpolated points (once the initial match is found) which would help to prevent falsely matching a segment when someone didn't actually cross a segment start or finish. With the current implementation there are a lot of opportunities to cheat or misuse segments. For example, one can simply turn around 30-50 meters before the segment end, and that segment would still match and that person could cut 10+ seconds if they are running. Perhaps it isn't easy to turn around on a bike but very easy when running, and I see that happening a lot on uphill segments.

By the way, I don't know if the same 500m limit applies to running segments. If that is the case that would really hurt uphill segments and stairs. There aren't that many steep hills that are longer than 500 meters.

What Strava does with increasing the limit has nothing to do with improving the absolute accuracy of segment matching. That merely reduces the relative error which makes the timing errors less noticeable.

If Strava engineers are interested, I could suggest an interpolation algorithm that wouldn't significantly increase the computing cost of matching segments.