Saturday, May 31, 2014

Changes to the analysis of the GPS accuracy of running watches

I’ve made two changes to my code for analyzing the GPS accuracy of running watches.
  1. My code was using the distance recorded for the first and last point in a lap, which should be the overall distance of the lap. However, for some watches the lap points do not reflect the complete lap, so I’m now using the distance recorded for the lap itself. 
  2. I saw at least one file from the 620 where GPS data was missing for the first couple of miles, even though distance data was recorded. I’m now ignoring those situations as anomalies.
This has a number of impacts
  • The accuracy of some devices improves slightly.
  • The ranking changes slightly, with the Garmin 620 and 10 improving while the Fenix 2 drops below them.
  • My testing of two different 620 watches now shows a greater difference between them. The original 620 now looks poor, but not terrible while the replacement 620 is far worse. This might explain why some people are seeing better GPS accuracy than others from the 620.