Results from 26 Runners' GPS units at the Run to the Beat Half Marathon
I was interested to read a
discussion between runners of a half marathon
held on 5 October 2008. 26 runners runners posted their GPS readings claiming
the course was long. However we know the course was measured by Hugh Jones
the IAAF/AIMS adminstrator for course measurement in English speaking Europe
and Africa, so we know that it will be the proper half marathon distance
ie 13.1091 miles+ 0.0131(Short Course Prevention Factor)
+/- 0.0131(maximum error for measurments with a calibrated bicycle and
a Jones Counter).
So the longest the course could be would be 13.1353 miles or 42 m longer
than the nominal 21.0975 km for the half marathon. n.b. measurers include
the SCPF to ensure the course will not be less than the advertised distance.
Here is a plot I have made from the 26 runners who posted their GPS results:
average distance measured was 13.39 miles - ( 0.26 miles or 2.2% long), the
results are clustered around this average with a standard deviation of 0.11
miles.( or 0.8%)
Compared with course measurers who would typically get results clustering
around the true distance with a standard deviation of about 0.004 miles (
0.03% ) we can see that the scatter of the GPS is about 25 times greater
than the measurers' calibrated bicycles. But in fact more worrying than
the scatter is the bias towards long measurement values of 2.2% which is
about 50 times worse than the typical bias for a group of calibrated bicycles
(which is actually derived from the accuracy of the precision steel tape
they use to measure their calibration courses and other sources.
Converting the above numbers to metric distances: course measurers would
get 21097m + 21m (ie SCPF) with a standard deviation of
7m and a bias of about +/- 8m. Whereas the runners' GPS gave a standard deviation
of 180m and a bias of + 460 m.
We can conclude that these runners' GPS results were 25 to 50 times worse
than the accuracy for an official course measurer.
What are the reasons for the poor GPS performance?
- Not starting and stopping the GPS at the course
start and finish
- Not following the Shortest Possible Route
- Loss of satellite visibilty under trees and in the shadow
- Imperfect algorithms in consumer type GPS units worn by
- Organiser laying out course longer than that measurered
by the course measurer
Most runners concentrated on point 5 - blaming the organiser. However experience
elsewhere shows that runners' GPS nearly always give long results eg my write
up of the Keevil 10k investigation. Normally
runners can check whether the course follows that the course measurer's
route by refering to the maps that are published on this website for all
races in the South of England that have a Certificate.of Course Accuracy
issued by me. However, in this case although I got a measurement report from
Hugh Jones and it checked out OK, the organisers of the race decided not
to proceed with obtaining a UKA road licence so they can not be issued with
a UKA Certificate of Course Accuracy. Therefore the the services of this
website for the publication of the course map are not available for this
race at present.
©Mike Sandford - 25 October 2008