Last week I was early to a ride, so I decided to do some laps around the the parking lot, figuring it would make a neat Strava segment that I could KOM for a day. When my friend arrived, he saw what I was doing and decided to do the same, which naturally destroyed my chances for the KOM. Even staring defeat in the face, I still went ahead and created the segment, but the results from Strava were truly shocking. We all know that there’s a lot of fudge involved, but I had no idea it was this broken.
Looking at the image above we can see the original segment on the top left. Oddly, even though I defined it as the number of times I went around the building (a hot dog place), Strava ended up breaking my definition in half. Weird, but no problem. The really wonky stuff is everything but the effort on the bottom right. What is going on there Strava?
Some would say that GPS isn’t precise enough for this job, and I would wholeheartedly agree, it is what we have though, and that doesn’t mean that we can’t build an algorithm that does some basic sanity checking (e.g. is the average speed greater than the maximum speed). Sanity checking that is clearly lacking from whatever Strava is doing; which appears to include anyone who has passed by, even in the most random way.
Now, just to be clear, the point of this post isn’t to complain that I didn’t get the KOM. That was never going to happen once my friend joined in. The point is that now that Strava is selling access to segments, it’s important to know what you’re buying. If they’re handing out titles based on what is clearly a poor guess, using imprecise data (the current leader has an average of 64.5mph, on a bike, around a three point intersection, with three stoplights, in a residential area, on a flat road; i.e. that’s not possible), with non-existent data checks, that significantly changes the value proposition. When it was free, they could make the get-what-you-pay-for argument in good conscience. Not sure that’s valid anymore.