Surprised that I achieved a KoM on a short downhill segment recently I wondered how accurate segment times are. My gut feeling had been that they are probably a few seconds out at each end, therefore potentially +/- 6 seconds error.

I recorded two rides each with the same two gps systems, one a phone app, the other an eTrex 30. In all 98 segments were recorded and the times generated were analysed.

11 of the segments were given identical times, a further 27 were within 2 seconds of each other, and 25 between 3 and 5 seconds out. That left 35 segments (36%) with a difference of greater than 5 seconds of which 7% were out by more than 10 seconds. The greatest difference was 19″.

I found no dependence on the length of the segment, which is not surprising as the error is most likely due to the start and finish positions, therefore two possible errors per segment of +/- E. For example if E=5 then a positive error of 5s at one end and a negative error of 5s at the other end would result in a total error of 10s as the maximum possible error. Then, when comparing two times for the same segment, if one has a maximum positive error of 10s and the other a maximum negative error of 10s we would see a difference between the two records as 20s.

What does this mean for Strava KoM records?

It means they are meaningless!

Take one of my KoMs, 10 riders are within 3s of my time, from my analysis each of those has a 60% chance of being 3 or more seconds “out”, or 3 (30%) therefore will have been faster than I was. Furthermore 147 riders were within 10s of my time, so approximately 10% or those will also have been faster than me. That puts me well out of the top 10 on the leaderboard for that segment. Of course I may well have been the fastest, but unless the time difference between me and the rest of the riders is at least 10s it is unlikely.

When we add this error to the difference wind direction makes, plus riding on your own versus in a group, it makes the KoM leaderboard meaningless.

Segment times however are fun, and taken as average measurements they can indicate increasing fitness, particularly for longer segments.