As you all know I have done a number of GPS tests for accuracy. Most of my tests have been with my bike mounted Garmin 60csx, but I've also done a few tests with a friend's Garmin Forerunner 205. My tests with the 60csx show an average difference of about 0.2% when compared to Jones counter measurements, with the error never exceeding 0.4% in any test. Three tests done while wearing the 205 while biking a course showed only slightly higher differences.
Scott Hubbard suggested that I should wear the 205 to see what effect the running motion has on results. I did this for one test and got a difference of 0.24%, which was in the middle of the three other tests I did with the 205, all performed while biking. Up until now, that was the only test I performed while running with a 205.
This all leads to my next experiment. I help organize a corporate competition that includes 5k and 10k road runs. I measured the courses for these two runs, but since they weren't being USATF certified I did not follow all the usual certification requirements. More on that later.
Before the race all the runners were asked to report their GPS measurements of the course to me, if they were wearing a GPS of course. 12 runners reported their measurements in the 10k and 11 in the 5k. 6 people reported in both races. The results are shown below.
After seeing these results I worried about my original measurement. Because there was no good place to lay out a long calibration course, I used a 100m course and recorded my measurement with the Jones. I decided to remeasure both courses, and this time I managed to find a spot for a 160m cal course and used rim readings rather than Jones counts when calibrating. The remeasurement of the courses resulted in:
5k - 4987 meters
10k - 9996 meters
During the measurements I also measured the courses with my bike-mounted 60csx. I also ran the 5k course while wearing a Garmin 205 and carefully following the SPR. The charts below are adjusted for the new Jones counter measurements of the course and also include the results of the 60csx and my run with the 205.
I have not yet run the 10k course with the 205, but after I do I will update the chart above.
Part of the error seen in the "runner-held" GPS measurements can be explained by runners not following the SPR, but it seems unlikely that is the source of all the error, especially since my 5k run, which did follow the SPR, was still more than 60 meters over the Jones measurement.
This is significantly more error than I've seen with the 205 while wearing it when riding my bike. It's possible the Forerunner GPS's are not as accurate when they are worn while running. All of this has convinced me to do more tests of the Forerunner while running, since I still have only two such measurements when I know the SPR was followed.
But regardless of whether it is caused by not following the SPR or by GPS inaccuracy, it is clear that it is quite common for runners to get GPS results that are 1-2% longer than the race course they are measuring.
This is nice work! Everybody's got an opinion, but it's work like yours that gets at the truth.
I measured a half marathon last year. the race was about a month ago, and almost immediately after it I got a forwarded e-mail note from one of the runners to the race director saying that her GPS had the course at 13.3 miles.
I sent them a note back explaining why that happened, and the race director let the runner know they had confidence in my work. She replied that she would be back next year but possibly without the GPS.
0.2 miles is approximately 1.5% of a half marathon...
This looks like a very good set of test results which is exactly what is needed.
I have found that the GPS distance accuracy improves with increasing speed. This makes sense to me since the positional accuracy of a GPS is typically 5m which means that the GPS may think it is within a 5m radius of its actual position. If the GPS is moving in a straight line the GPS may "think" it is moving in a zig zag line, because at one moment it may "think" it is to the right of the straight line and a moment later "think" it is to the left. This zig zag line will cause the distance measurement to be slightly longer than the straight line. As speed increases the zig zag straightens out and the distance accuracy increases. So walking is less accurate than running and running is less accurage than biking. Also, at my age most runners will have a higher accuracy than me.
Another difference between a wrist mounted GPS used for running and a handlebar mounted GPS used for biking is that on a bike there is a clearer view of the sky. When running the body, including the arm will block some of the satellite signals.
AIMS/IAAF "A" Measurer
If it could be coordinated, it would prove interesting to get a group of about 5 GPS wearers to run the same uncertified race course and report their findings.
Or, lacking numbers, get one GPS wearer to run 3-5 uncertified race courses and report their findings.
It would take additional doing but we'd need to get the uncertified courses measured (before or afterward) per cert standards so we could learn actual length.
Scott, I think that's what I did.
Just anecdotal, but this spring I ran w/ 2 friends who were both wearing identical Garmin Forerunners. We all ran the same course, staying on virtually the same path. At the end, their readings differed by almost a tenth over a 4.2 mile (uncertified) course.
Also, I seem to recall Norrie Williamson doing some experiments where he wore the GPS in different ways (wrist, upper arm, even on his head!) and the movement of the body part seemed to have an adverse effect on accuracy.
Here's an update. Since my original post I have:
1) Had someone run the 10k course wearing a GPS and carefully following the SPR.
2) Reran the 5k course wearing a GPS and carefully following the SPR.
3) Remeasured the 5k. This remeasurement was within a meter of my last one, at 4986 meters. I am now convinced that this is very close to the true distance of the course.
Below are my charts with these new results.
That first GPS measurement I got from a Forerunner was the first and only one I have gotten that was more than 0.4% different from my Jones counter measurement. That's why I ran the course again, and this time I got a result that was more in line with all my other GPS results, with about 0.32% error.
I think most of the reason the runners in races get long readings is because they do not follow the SPR. This is often a good idea. The 10k course in this case had a turn-around and both courses had several 90-degree turns. Adding a few meters to the course by rounding out those turns is almost certainly faster than trying to strictly follow the SPR.
But there are also some winding sections on the course, and I doubt many runners followed the SPR here either. I know that when I was racing, before I started measuring, I didn't pay a whole lot of attention to looking ahead to the next tangent line. Now, like with all course measurers, this has become second-nature to me.
My conclusions from this study are:
1) Runners wearing GPS devices in races will usually get measurements that are more than 1% longer than the actual course length.
2) Most of the error in these measurements is not due to the GPS device itself, but due to the fact that the runners do not follow the SPR, and/or start their watch before reaching the start line, and/or stop their watch after crossing the finish line (I saw a few of them do that).
3) There were 4 different models of the Garmin Forerunner used by runners, 201, 205, 301, and 305. The average difference between any of the models was at most about 0.2%.This message has been edited. Last edited by: Mark Neal,
I am often managing the course from a bicycle, and while riding along find it frustrating to see runners putting a lot of effort into running what amounts to the wrong way.
The most noticeable place this happens is when they run the wrong way in a triathlon transition zone: That is truly the wrong way. But I see the more subtitle version in running races, when people fail to run the shortest route to the finish.
If I am the lead bike I will try very hard to ride the shortest path, because without that guidance even the elite runners at the front are off track. Elite runners are some what better at it than most, but they still wonder all over the place. I think that's becuase in their minds they are not running against the course but the other runners.
For the rest of the pack it gets worse. I often most of the runners taking the corners wide, or runners running on one side of the stream of people, instead of crossing the stream diagonally to be on the inside of the next turn. This is partly because they cant see far enough ahead to plan the route, and partly because they have trained themselves to run looking down at an angle, so the focus of their gauze is below the horizon. While this gives a good view of the t-shirt in front, or the bums beoyond, it's not good for navigation.
I have overtaken people in races because I was on the inside of a curve and they were on the outside.
It amazes me that with all the stuff written about running and training, the simple act of looking up and navigating the shortest route is rarely mentioned.
Maybe I'm more aware of direction because I grew up racing small sailing dingys, where you are constantly thinking about navigation, the direction you are pointing and making the fastest time to the next mark. In sailboat racing, like track running, you fight to get the inside position on every turn.
It's a shame more runners are not navigationaly aware and look up to figure out the next tangent.
I have recently acquired a Garmin Etrex H which I have been testing on an accurately measured 4.5 km course. I am at present analysing the data, and I should like to compare it with your results. Do you have readily available more detail of your results from your bike rides with your Garmin 60cx. How many rides did you do? What settings did you have on your Garmin, and how did you take the results? From the odometer? I understand your deviation was +0.2%. Was this a mean deviation? What was the standard deviation of your results?
So far I have had 7 rides with the average differing from the actual course length by +0.02%, and the standard deviation is 0.17%
Mike Sandford -
Measurement Secretary South of England
UK Asssociation of Course Measurers
email contact m.sandford at lineone dot net
I did my tests a bit differently. Rather than measure a course first with the Jones and then later measure it with the GPS, I measured it with both at the same time. I simply calibrated my bike and then went for a ride of arbitrary length. This approach had the following advantages:
1. It measured only the difference between the two types of measurements, since the results were not confounded by differences in the paths I would follow in different rides of the same course.
2. It made the tests much easier to conduct since I didn't need to follow a particular course or worry about the SPR, or for that matter, worry about much of anything during my rides.
3. Since my GPS would show measurements only to the 100th of a kilometer, at the end of each ride I would slowly roll forward until the next 100th of a kilometer turned over. This gave me a GPS measurement down to something close to 1 meter. This wouldn't have been possible if I had measured a course of fixed length.
You can download an EXCEL file of the results of all my rides at
I've also put the EXCEL file of my other experiment with runners in a race at
Thanks for the data.
I like your method of squeezing out more precision than is provided by the digitisation to 10m intervals on the odometer. I think if I mark the start/finish of my loop with some 1m interval distance markers, I can follow your method.
My loop has been very accurately measured by Jones counters and with a steel tape. I have little difficulty in following the SPR after many many rides over it!
The numbers I quoted above come from post-processing of the downloaded tracks, in order to smooth out the sample fluctuations when set to record 1 sample per second. I want to compare this method with the odometer reading and I obviously need more rides looking for the roll over.
I agree with your overall conclusion that these devices can be remarkably accurate when mounted on a bike and ridden along the same path that is measured by the Jones counter.
From your data it looks as if it may be possible that some older models of GPSs worn by runners may be less accurate.
Mike Sandford -
Measurement Secretary South of England
UK Asssociation of Course Measurers
email contact m.sandford at lineone dot net
Yeah, multiple rides of the same course are going to be different by only a couple meters. The main reason I did it the way I did was to make it easy.
The runners in my other test were wearing four different Forerunner models I think. There wasn't an obvious difference in the errors of the different models. I think the improvements over the last few years have mainly been that the newer ones are able to get and hold onto a signal better. But I don't think holding a signal was an issue on this course.
The big surprise for me in that second test was how much extra people are running in races. The average was over 10 meters per kilometer!
|Powered by Social Strata|