In GPS We Trust: Runner threatens law suit over course measurement disagreement.

Not sure if this has been mentioned, but these courses are rarely "measured" and "certified" on race day. So although at one time an accurate measurement of a course has been made, the chances that you are running that same course on race day are not 100%. Ever been to a race where they simply got the start and finish lines mixed up? Turn around cone placed incorrectly? These things can happen despite everyone's best efforts. No idea if they played a factor in this case or not. I ran a USATF certified race once where they had to move the start line for safety reasons. The race wound up being about 400 meters short. After a week or so of deciding what to do the USATF announced that the finish times would count as "certified 5k times".

dukerdog wrote:

His statements that the GPS units calculate distance by simply "connecting the dots" and that the GPS units take measurements every 1 second or 5 seconds, are things that he doesn't know. The units may STORE points only every 1 second, but for all we know they are taking measurements every 10 milliseconds and using those to help calculate distance. And they may be applying real-time filtering to the points as part of the process of calculating the distance, not just connecting the dots. We just don't know. To state these two things as limitations of how accurate the units can be is wrong.

measure the distance of each segment (between the "dots") and sum them. if the sum of segments equals the total reported distance, then there are either actually, or effectively, no intermediate position measurements. the faster you move around turns, the easier it will be to tell what is happening.

rixctf wrote:

measure the distance of each segment (between the "dots") and sum them. if the sum of segments equals the total reported distance, then there are either actually, or effectively, no intermediate position measurements. the faster you move around turns, the easier it will be to tell what is happening.

I have done this before by downloading my track to google earth, checking the distance calculated by GE, and comparing it to the distance reported by my GPS. I got somewhat different numbers. Is that due to GE inaccuracies or Garmin not simply connecting the dots, I don't know.

dukerdog wrote:

I have done this before by downloading my track to google earth, checking the distance calculated by GE, and comparing it to the distance reported by my GPS. I got somewhat different numbers. Is that due to GE inaccuracies or Garmin not simply connecting the dots, I don't know.

Maybe try having two of the same gps devices on two bikers, runners, whatever. Have them run the same path at the same time side-by-side, over a known distance. Preferably straight I guess. After ~100m, have one of them turn. If there is error in the one that turned, but not in the one that went straight, this may indicate the error is in the integrated path measurement, and not in the instantaneous location measurement. This assumes that the triangulation error will be the same for identical devices within a few feet of each other.

How many times would you need to stop and record the click count with the Jones unit over 33 plus miles? Or do you just keep track of how many times it rolls over?

Yepper wrote:

How many times would you need to stop and record the click count with the Jones unit over 33 plus miles? Or do you just keep track of how many times it rolls over?

None, a 6 digit counter will do 50 miles without a rollover. (or one rollover, depending on the start number).

dukerdog wrote:

I have done this before by downloading my track to google earth, checking the distance calculated by GE, and comparing it to the distance reported by my GPS. I got somewhat different numbers. Is that due to GE inaccuracies or Garmin not simply connecting the dots, I don't know.

But it doesn't seem that you compared the sum of all the segments to the GPS's reported total. He/she wasn't suggesting you compare reported distance to an outside measuring source but to the GPS itself. This would tell you if there was some sort of correction factor or accurate smoothing action (whatever you would call it) as you seem to suggest.

If the sum of all the segments adds up to the GPS reported distance then the accuracy of the GPS is clearly specious. Especially when, for example, the map around the track shows detours into the bleachers (I see this on nearly every GPS map I have ever produced).

So does the GPS just add up this segments... you don't seem to know.

Eric B. wrote:

dukerdog wrote:

But it doesn't seem that you compared the sum of all the segments to the GPS's reported total. He/she wasn't suggesting you compare reported distance to an outside measuring source but to the GPS itself. This would tell you if there was some sort of correction factor or accurate smoothing action (whatever you would call it) as you seem to suggest.

If the sum of all the segments adds up to the GPS reported distance then the accuracy of the GPS is clearly specious. Especially when, for example, the map around the track shows detours into the bleachers (I see this on nearly every GPS map I have ever produced).

So does the GPS just add up this segments... you don't seem to know.

Agreed. I didn't read the post clearly the first time. Sum the segments. I want you to compare GPS segments to GPS total. This will tell you if the GPS location is measured only at the endpoints of segments, or if, as you suggest, position is measured more frequently and the segments merely represent a data storage issue.

I would like a GPS watch that lets you first input the run you want to do by using a program like google earth and mapping out your run. Then you download that map into the watch and start the run. Now the watch has a reference point for your intended path from which to calculate a more accurate speed and distance.

I use GPS extensively in all my training and urge all my athletes to do so.

rixctf wrote:

Sum the segments. I want you to compare GPS segments to GPS total. This will tell you if the GPS location is measured only at the endpoints of segments, or if, as you suggest, position is measured more frequently and the segments merely represent a data storage issue.

Not that it can't be done, but depending on the format of the track data coming out of the GPS, "adding up the segments" probably isn't a trivial task. Likely it involves trigonometry in three dimensions and an understanding of the altitude reference frame. By importing the data into Google Earth or other such tool he's using someone else's canned solution to this problem. Unfortunately you don't know how how Google Earth manipulated the data either.

I have been measuring for USATF certifications for 5 years, and over 150 courses. I have ridden with a GPS (Garmin eTrex Vista, Venture, and latest - Oregon 450). Rarely do I get a GPS track the same length as the calibrated Jones counter.

I have exported the data as a txt file, and added the segments. The totaled segments match the displayed distance for each track. I have set my units to record a point as often as possible, using the distance method. This allows a trackpoint every 1/100 of a mile, or 52 feet. The segments are anywhere from 40 to 65 feet in length, so even when set to measure every 52 feet, it varies. Sharp turns are rarely tracked properly. I have also set them (when I am not measuring a race course) to measure every 5 seconds, or 1 second, and the tracks are not smooth. Straight lines from one point to the next, but very jagged, and not in the straight line that was ridden.

I have run 3 miles with one unit in my left hand, and one unit in my right hand. Being ambidextrous with GPS units (are you impressed???), I started the tracking simultaneously, and ended them simultaneously. The tracks deviated from each other along the route, with one venturing to the other side of the street on one occasion, but rarely were they parallel for more than 50 feet. They also were 64 feet different in length.

This, plus other tests and results from my calibrated rides, prove to me that a GPS unit is not accurate enough to state that a course is the wrong length. Especially when you consider the test in the previous paragraph.

Dukerdog, you had a link to a track that was produced by your wife, in Detroit, I believe. That track contradicts your statement about "collie" behavior. It clearly shows she ran through some buildings along the course. I see this behavior on every course I measure that is between buildings over 5 stories tall.

Keep in mind that we never hear from the runners who show 13.11 miles when they have run a Half-marathon, or 26.22 miles when they run a Marathon. More people likely show the correct length than don't, I imagine. Also, just because a GPS device CAN be inaccurate, doesn't mean it ALWAYS is. Even a broken clock is correct twice a day. Some users experience incorrect readings every time. I invite them to go along on a measurement with a USATF measurer, and see the process. Bring your GPS. We can show you where you are possibly inaccurate, or point out the trees and buildings that may be causing you problems.

This lawsuit has no merit, if the course is shown to have been properly measured (a validation ride will verify accuracy).

RaceMeasure wrote:

Dukerdog, you had a link to a track that was produced by your wife, in Detroit, I believe. That track contradicts your statement about "collie" behavior. It clearly shows she ran through some buildings along the course.

Ouch! Really, there is no reason to bring his wife into this. Low blow.

RaceMeasure wrote:

This, plus other tests and results from my calibrated rides, prove to me that a GPS unit is not accurate enough to state that a course is the wrong length. Especially when you consider the test in the previous paragraph.

This statement is too general. Yes it is correct for a certified course. But there are two occasions that a GPS is accurate enough to say that a course, or some aspect of it, is significantly off:

1) misplaced mile markers on a road course. These can sometimes be much further than 1% in error. Even a simple watch will suggest this (wow that first mile was slow!)

2) trail courses. GPS watches are even less accurate on a trail than on a road, although on a trail they will often underestimate the distance because they do not capture the turns very well. But gps distances on many, many trail courses are far, far off from the advertised distance. Who knows what the true distance is or how the RDs estimated it. Indeed I would even argue that the true distance is fairly meaningless for many trail races.

I am surprised at the number of people who think a handheld GPS units is an accurate way to measure a course. I work a lot with professional GPS equipment and have occasionally worked with land surveying equipment. When measuring a point the difference between the two is signficant (usually under a meter), and even more so when compared to the consumer level GPS units (5 meters). It gets even worst when measuring a line.

There are so many factors involved with hand-held units that makes it really difficult to be highly accurate. The accuracy of a hand-held unit is dependent on the number of satellites, signal strength, data quality, satellite calibration, software rounding, processor capacity. There are too many possible errors to be considered accurate, and would not be considered a legal measurement.

.....Case dismissed.....

I "brought his wife into it" only because she wore the GPS in Detroit, and the track was presented as evidence of GPS accuracy. Doesn't matter to me if Jack the Ripper wore it, I was just referencing the track itself.

A GPS is accurate enough to raise suspicion that a course may have been run/laid-out incorrectly, but not to flat-out state that a course is long, just because someone's GPS said 13.12 or 13.13 miles for a Half. GPS-wearers felt a course needed to be checked a month ago, and it turns out that a turnaround was in the wrong spot. However, it was 768 feet too long, not just a hundredth or two. My intent was to say a GPS is not accurate enough to lodge a complaint if it is within two hundredths, which seems to be the most-common complaint. Look into it, sure, but don't automatically assume the course was measured incorrectly.

As with all allegations, more than one piece of evidence needs to be examined before filing a lawsuit. A GPS is only one piece of evidence.

Citizen Runner wrote:

Not that it can't be done, but depending on the format of the track data coming out of the GPS, "adding up the segments" probably isn't a trivial task. Likely it involves trigonometry in three dimensions and an understanding of the altitude reference frame. By importing the data into Google Earth or other such tool he's using someone else's canned solution to this problem. Unfortunately you don't know how how Google Earth manipulated the data either.

You're exactly right. It's not a trivial task to calculate the distance between two GPS coordinates. We don't know what formulas Google uses to do this, but we do know the only data they are using to do it is the track points reported by the GPS, since that's all you give them.

Eric B. wrote

So does the GPS just add up this segments... you don't seem to know.

That was the whole point. I don't know. And neither does anyone else, except Garmin.

RaceMeasure wrote

Dukerdog, you had a link to a track that was produced by your wife, in Detroit, I believe. That track contradicts your statement about "collie" behavior. It clearly shows she ran through some buildings along the course. I see this behavior on every course I measure that is between buildings over 5 stories tall.

Me too. That's why I said this:

dukerdog wrote

All of this goes out the window when there are tall buildings around. In that case you will see crazy tracks and bad distance info. But he is suggesting it happens all the time.

RaceMeasure (since we're using letsrun names ;-) ) Why don't you and Keith both do a careful experiment to compare your Jones counter to your GPS. Don't do this while measuring a course for certification. You have too much else to worry about while you're doing that to also worry about doing this test correctly. Every measurer who has done this (4 of them) and reported the data on that other forum has reported differences of less than 0.5%.

dukerdog wrote:

It's not a trivial task to calculate the distance between two GPS coordinates.

Believe it or not, this part is actually a very simple calculation. It's basically the Pythagorean theorem.

Big John wrote:

dukerdog wrote:

It's not a trivial task to calculate the distance between two GPS coordinates.

Believe it or not, this part is actually a very simple calculation. It's basically the Pythagorean theorem.

I guess I might elaborate though and say that although it is easy to compute the (shortest) distance between a pair of GPS coordinates, that is not exactly the "real" distance traveled by the receiver. For one there will be some error in each coordinate measurement, and then there is the issue that the receiver may have not followed a straight line between coordinate measurements.

I was in the GPS game about 10 years ago developing military grade hardware. I don't know what chips Garmin uses or what they do with the data during computation, but I'd stake my Letsrun.com reputation (LOL!!!!) that it's a Kalman filter :)