www.gmap-pedometer.com

Hey, even at just 0.5% it adds up. Throw in poorly tangented curves and you'll easily get 5% error.

It's very accurate. I tested it by measuring the length of the track at Kezar Stadium in SF. It said 400.4 meters. The .4 is easily due to the tangential measurement of the turns.

MarathonMind wrote:

Hey, even at just 0.5% it adds up. Throw in poorly tangented curves and you'll easily get 5% error.

No, you obviously don't understand the math. What "0.5%" errors are you adding up?

If you were running up a hill that gained 200 feet in a horizontal distance of 5280 feet, you would have an error of about 0.07%.

If you then ran down the other side of the hill (exact same geometry), your total error is still 0.07%.

Go up and down the next hill (exactly the same geometry) and your total error remains 0.07%.

Do it again and by now you've covered 6 miles with an error of: 0.07% (not 6 x 0.07 = 0.42%).

As for poorly tangented curves, that has nothing to do with your original post about elevation changes and is entirely the responsibility of the user, not the lack of a third dimension on the mapping software.

If you are 'easily' getting 5% error with your curves, then you are simply not very good at using the software (or are using very poor resolution imagery).

Yes I understand the math. Percentages are constant.

MarathonMind wrote:

Yes I understand the math. Percentages are constant.

Huh? Then why the bit about things adding up? All we have to do is consider the elevation changes of a given course, add up the magnitudes, and then divide by the length of the course to determine the average grade for the purpose of error correction. No reasonable course is going to be anywhere near 10% and as already indicated, a 10% grade produces only .5% error. dukerdog was very explicit at the outset that other sources of error will dominate and Asterix rightly points out that poor mapping is not a problem of the tool, but of the user of the tool.

When Google Earth first came out I cross-checked it a number of ways, including zooming in on a couple of local tracks with slightly different geometries. Generally, I was no more than +/- 1 meter off out of 400 for +/- .25% deviation. Further, sometimes I was slightly over and sometimes slightly under, so although the worst case magnitude was around .25%, the average of several trials was better.

Finally, I'd rather have the tool tell me the course is shorter than it really is (which is the case here) than the opposite, especially if I'm cross-checking for eventual certification.

Bottom line: There is absolutely no need to worry about elevation-induced errors when using Google Earth on any reasonable running route.

Bottom line: There will be errors on the short side, owing to elevation change alone.

Also, I have found errors in Goole Earth where adjacent photos are not married correctly.

MarathonMind wrote:

Bottom line: There will be errors on the short side, owing to elevation change alone.

But do the math. Any error on the short side is going to be 0.5% or less (ie a 10km could be off by no more than 50m, or 10 - 15 seconds).

You're far more likely to have errors in your following your exact route, so elevation change is pretty much irrelevant.

Also, I have found errors in Goole Earth where adjacent photos are not married correctly.

That's true too and could have an effect. Until all states recognize same-sex weddings, not every marriage will be correct.

JimFiore wrote:

When Google Earth first came out I cross-checked it a number of ways, including zooming in on a couple of local tracks with slightly different geometries. Generally, I was no more than +/- 1 meter off out of 400 for +/- .25% deviation. Further, sometimes I was slightly over and sometimes slightly under, so although the worst case magnitude was around .25%, the average of several trials was better.

But as I'm sure you know Jim, the problem with that technique is that you know what the correct answer is ahead of time. You really need a blind test.

A while ago I put a map of a course on this forum and asked people to measure it with whatever mapping program they chose, and let them submit their answers to me anonomously. I had already measured the course for certification, so I knew what distance the calibrated bicycle method gave me.

With all the questions on here about the accuracy of the mapping programs, I was surprised that I only got two submissions. Both those submissions were about 0.5% different from my calibrated bicycle result, and that agreed with what I had gotten myself with the mapping programs. But of course, with only two data points it's difficult to draw any conclusions.

I'd be happy to do it again with a different course but I'm guessing I'd get a similar response.

dukerdog, while it's true that I know the answer (400m) beforehand, I don't know my map answer until it's done (I just push the path window out of the way so I can't see my progress). It's not really double-blind but it helps to keep me honest. Point taken though.

If we're talking about errors due to elevation, what we need is to determine the total magnitude of elevation change of typical courses (or very hilly courses for worst case). As stated, 10% grade yields .5% error but I've never run a race that averaged out to 10% grade. I'd guess that even my most hilly training loops don't average that. That's 1000 meters of rise and fall over a 10k.