Testing the Accuracy of a Garmin GPS

I think the results would show that the accuracy of the GPS measurement, besides the obvious starting and stopping of recording the track at the precise Start and Finish lines, depends on the course.

If a course runs downtown, where there are skyscrapers around, the accuracy is greatly impacted. Same with running under a canopy of heavy tree cover. If there are many sharp turns, or turnaround points, that reduces accuracy, also.

I would love to get results from many units used in the same race, but apparently, many people can't download a track from their unit. I use a mapping unit, so that is not convenient to carry while racing (I no longer race). So, I will try to occasional solicit track files from an event, but I am not holding my breath.

Vic, there has been a good bit of testing done on gps units. If
You click on the Find tab and then searh for GPS, several threads
Will come up.

The other "test" of these units would be to create a track while actually measuring a course. You'd have to be going straight through, not parking the bike to paint a mile mark etc.

The few tracks I have done while measuring have shown anomalies, like the time I thought I was riding around Ohio Drive near Hains Point, and the gps thought I rode all over the golf course! Weird because no tall buildings and only a few small trees.

Bob, I do a GPS track with every course I measure. I lay it down when doing my straight-through ride.

While I have checked in the past and found that the length is normally not correct, I don't even look at the length of my GPS track any more. Too inconsistent to even consider.

I do notice, though, as I map my course using the GPS track as an underlay, sharp turns are not recorded accurately. I have mine to drop a trackpoint every 1/100 of a mile (the finest distance-related tracking it will do). It may do better if I tell it to drop a point every second, but that would be a much larger file. May not even be able to hold all the trackpoints in one file for a half-marathon course.

Seems like you need variable tracking instructions-- on turns or other tricky changes you need fine-tuned spacing, but on a long straight portion every 50 feet could be enough.

In the end, all that matters is the distance the GPS reports at the end of the ride. I've seen some cases where the stored track doesn't look all that good as far as following my true path around turns and such, and yet the final distance traveled that the GPS reports was very close (within 0.2-0.3%) to the Jones counter distance calculated on the same ride.

These units are almost certainly not calculating distance by simply "connecting the dots" on the track points that they have saved for you. It's much more likely they are taking position readings much more often internally and using that information to calculate the distance you traveled. Once they have done that to calculate the distance traveled for the first say, 5 seconds, of your trip, they can discard that data and start taking readings for the next 5 seconds. They can't save all those readings in your track log though because it would take take too much memory.

It's very easy to do a careful test of your GPS accuracy. Do a set of calibration rides to get the cal constant for your Jones. Then go for a ride with your GPS. You don't have to follow a certified course or follow the SPR, just ride. But no backing up, no getting off your bike, and no freezing your front wheel and walking. When you finish your ride calculate the distance based on the Jones counts for the ride and compare to the distance your GPS reports.

Have there been any tests done to show repeatability of a gps or between multiple gps?

I haven't tested that directly, mostly because it require that you carefully follow the same route multiple times. When you do tests to compare to the Jones counter, it doesn't matter what route you follow, since the Jones and the GPS are measuring the same route. Mike Sandford did do multiple rides of the same route with his GPS.

But you can certainly get an idea of the variability of GPS measurements from my tests, simply by checking the variability of the difference between the GPS and Jones measurements. In the small number of tests shown in this thread
https://measure.infopop.cc/eve/...9510622/m/8301003542
the range of difference was from -0.33% to +0.34%

I have also gone on a couple rides wearing my Forerunner and with another GPS mounted on my bike. The difference between the two was similar to what I saw among different rides.

BTW, the chart in that thread shows that you are just as likely to get a short measurement from a GPS as a long measurement. It's not true that GPS "always measures long." It IS true that runners in races wearing GPS always run as long or longer than the SPR, by definition, so their GPS measurements are almost always long.

Locally we all deal with these nagging, sometimes foolish, questions of course accuracy comparing a GPS watch or devise measurement to a certified course measurement.

USATF periodically issues public announcements on various subjects related to the sport. Why not an announcement related to differences between GPS measurements and what we do when measuring and certifying a road course?
The USATF announcement could be something we can point to for a simple explanation about the differences in measuring results between GPS and certified courses. More importantly, like many things, when a governing organization issues a statement, it carries much more weight and tends to dispel doubt. While bringing clarity to GPS vs. Certified Course argument, the announcement could also shed some light on the process of measuring, marking, documenting, and certifying a road course distance.

How many of you would find an official announcement on the GPS vs. Certified Course argument helpful?

I would vote for that, Kevin. Something not too long or too opinionated, but possibly could give references for further reading for someone who is interested?
Several of these have been written and they have been shared on the forum.

Agree.

A local race operator (Hartford Marathon Foundation) has a statement on their website explaining that GPS may not be accurate. Part of it is copied below.

1) Courses are measured by the shortest-possible route available to runners on race day. This means
that the measuring bicycle is ridden within a foot of the curb and turns are “straightened out” as much as
possible. This method is used to ensure that no one runs shorter than the stated distance. With other
runners on the course, however, this shortest route can be difficult to run exactly as measured.
2) All certified courses include a 1/10th of 1 percent (one-thousandth percent) "short course prevention
factor." This is a small extra cushion to again make sure no one runs less than the stated distance. So, in
other words, a certified 10 km (10,000 meters) road race is actually measured as 10,010 meters. For a full
marathon, that means 42.1925 extra meters or a little more than 135 feet. This extra cushion is spread out
throughout the course, not simply added to the start and finish, and is present in ALL certified courses.
3) And, lastly, consumer GPS devices are not 100% accurate. We have found they are typically 1-to-2%
off. GPS accuracy is affected by elevation, the number of turns, tree coverage, tall buildings, bridges and
overpasses, and the quality and quantity of satellite reception. In a half marathon, a 1% difference is more
than a tenth of a mile. So, readings of 13.25 miles or more for a 13.10938-mile certified half-marathon
and 26.50 miles or more for a certified 26.21876-mile marathon are not unusual and are to be expected.

I'm not sure I agree with #2, but it could be used to develop a simple statement that could be put on a certificate.

Statement #3 is not supported by any testing that I am aware of.
I agree that #2 is probably not necessary. Runners wearing GPS don't complain that a 10k is 10 meters long, they complain that it's 100 meters long (which is due primarily to #1). The SCPF is irrelevant as far as they're concerned.

Complaints I've seen are claiming way more overdistance than that, stuff like "26.7 or 26.8 miles". Just saying.

If a GPS is mounted on the wrist, and the runner runs the SPR, the path the GPS takes will be a zigzag as the arm swings around. When you are running your wrist does not follow the same path as your belly-button. It follows a longer path.

So, if the GPS is working perfectly, the path it takes will always be longer than the SPR.

I think this may complicate things when it comes to testing.

No part of a person's body ever goes backward relative to the ground while they are running. The wrist is always moving forward relative to the ground, or at worst, is stationary. So your wrist is not traveling further, it's just not traveling at a constant speed like your belly button.
I've also done tests where I have run with a GPS on my wrist, carefully following a course I measured with my bike. I didn't see much difference in accuracy when compared to a bike mounted GPS or a GPS I was wearing on my wrist while biking.

Not backward, but certainly side-to-side and up and down. The Jones Counter measured line is a lot straighter than the line traveled by the runner-worn GPS. That's got a series of small zigzags all along the way, all of which make the GPS traveled path a bit longer. How much difference these zigzags make is unknown to me.

How to duplicate the path of a runner-worn GPS in a test is quite a problem. Certainly the bike comparisons show the GPS to have, in many cases, good accuracy - but the bike tests don't replicate the path that a runner-worn GPS takes.

It's very unlikely that GPS devices calculate distance by simply connecting the dots and summing up the distance of the line segments. More likely, they collect a time window of location points and then signal process that series of points to determine the most likely path the person actually followed. Part of that signal processing is to remove high frequency content that is in directions perpendicular to the main direction of travel. So if you are swinging your arm side to side as you run, they can see that in the signal, and they know they shouldn't include that movement when they calculate the distance traveled.

Here's one test that may have already been done:

Calibrate and ride a bike on a straight course over as long a length as can be found. Read the GPS.

Strap on the GPS and run the same course.

Compare the two measurements.