Dogtrousers
Kilometre nibbler
This is something that we often discuss - how bad, or good, GPS units are at telling us the total elevation gain for a ride. GPS is optimised to give a location on the earth's surface, and not so good at measuring elevation. So we often get people complaining about inconsistent "total ascent" values from GPS units.
So I took 83 Strava uploads for stage 4 of this year's Tour. The first problem was that they were all different lengths, as some started recording in the neutral zone, some before, and some at the real depart. So I picked a 2 km interval 177-179 km and got 67 rides. They were mostly clustered around 178km. Given that the nominal distance for the stage was 171.5km I guess this is the départ fictif to the finish.
This was the same ride, at broadly the same time, and all riders will be using top of the range GPS units.
Then I looked at the distribution of the total elevation values.
Each bar shows the number of rides that recorded a value between the number shown at the bottom and the start value of the next bar.
The bottom line is that 94% of the rides are within plus or minus 104m of 1852m. Thats plus or minus about 6%. It's a skewed distribution which makes interpretation tricky for someone who last studied statistics 40 years ago.
This doesn't address a whole load of questions, such as:
Would I get the same variance if I did the same ride repeatedly (under the same meteorological conditions) with the same GPS unit, ie does it result from a variance between units or is it intrinsic imprecision of the GPS technology? (I suspect mainly the latter)
Is it better or worse to rely on "contour counting"? Either with pencil and paper or letting a system like Strava or RWGPS rely on its map database to do this for you. (I suspect that the question is meaningless)
But if you and your riding partner go for the same ride and your GPS units give different total ascent values with a difference of less than +/- 12% don't be very surprised. If the difference is less than +/- 6% don't be surprised at all.
Edit: In case anyone's interested, the three outliers to the left are: Jan Tratnik, Garmin Edge 830 (178.1km, 1332m); Krists Nielands, Hammerhead Karoo 2 (178km, 1649m); Franck Bonnamour, Garmin Edge 830 (177.7km, 1454m). The highest value was Aleksandr Riabushenko, Garmin Edge 530 (177.7km, 1956m)
So I took 83 Strava uploads for stage 4 of this year's Tour. The first problem was that they were all different lengths, as some started recording in the neutral zone, some before, and some at the real depart. So I picked a 2 km interval 177-179 km and got 67 rides. They were mostly clustered around 178km. Given that the nominal distance for the stage was 171.5km I guess this is the départ fictif to the finish.
This was the same ride, at broadly the same time, and all riders will be using top of the range GPS units.
Then I looked at the distribution of the total elevation values.
Each bar shows the number of rides that recorded a value between the number shown at the bottom and the start value of the next bar.
The bottom line is that 94% of the rides are within plus or minus 104m of 1852m. Thats plus or minus about 6%. It's a skewed distribution which makes interpretation tricky for someone who last studied statistics 40 years ago.
This doesn't address a whole load of questions, such as:
Would I get the same variance if I did the same ride repeatedly (under the same meteorological conditions) with the same GPS unit, ie does it result from a variance between units or is it intrinsic imprecision of the GPS technology? (I suspect mainly the latter)
Is it better or worse to rely on "contour counting"? Either with pencil and paper or letting a system like Strava or RWGPS rely on its map database to do this for you. (I suspect that the question is meaningless)
But if you and your riding partner go for the same ride and your GPS units give different total ascent values with a difference of less than +/- 12% don't be very surprised. If the difference is less than +/- 6% don't be surprised at all.
Edit: In case anyone's interested, the three outliers to the left are: Jan Tratnik, Garmin Edge 830 (178.1km, 1332m); Krists Nielands, Hammerhead Karoo 2 (178km, 1649m); Franck Bonnamour, Garmin Edge 830 (177.7km, 1454m). The highest value was Aleksandr Riabushenko, Garmin Edge 530 (177.7km, 1956m)
Last edited: