Accurate distance: I calibrate my SRM Power Control head unit by measuring distance along a long measuring tape using at least five revolutions at a specific tire pressure while my weight is on the bike as I would ride. I then divide to get a circumference good to a millimeter or so (should be ~0.05% error or so). This delivers accuracy down to 1/1000 of a mile (5 feet) using the SRM. If I ride the same course over and over, I get the same figure every time to 0.001 mile. If I change tire size (different wheel), I can immediately see the error in distance over the same course (unless I recalibrate).
I do not use GPS for any of my cycling, having found it highly inaccurate in both distance and a bad joke as far as altitude (losing 1000' all of a sudden, reading descent while ascending steeply, etc). GPS such as the Garmin Edge 500 produces garbage data under the conditions I would most need it (mountain biking with tree cover and switchbacks, but also road biking under tree cover).
Some folks ride flat open roads, and there GPS is presumably not too bad. But all those KOM things... well now IEEE Spectrum reports on GPS inaccuracy in Why Every GPS Overestimates Distance Traveled.
Runners, mariners, airmen, and wilderness trekkers beware: Your global positioning system (GPS) is flattering you, telling you that you have run, sailed, flown, or walked significantly farther than you actually have. And it’s not the GPS’s fault, or yours.
Blame the statistics of measurement. Researchers at the University of Salzburg (UoS), Salzburg Forschungsgesellchaft (SFG), and the Delft University of Technology have done the math to prove that the distance measured by GPS over a straight line will, on average, exceed the actual distance traveled.
As an example of “lab testing blinders”, the researchers (mathematicians) quoted in the article seem to be clueless about far larger errors in the real world! Their findings are laughable in the face of actual real-world errors in the opposite direction of error (too short in distance)—real world issues that relegate their findings to a rounding error!
It’s a classic case of not seeing the forest for a few sticks of wood. Their “test” consisted of having subjects walk a 10 X 10 meter square. While this validates the mathematical model (apparently), how is that realistic for anyone anywhere using a GPS to track distance? Do you walk in a 10 X 10 square, or have you ever?! Science like this which uses a context bearing no relation to actual usage conditions is devoid of knowledge.
So here is some knowledge of actual GPS performance based on real-world usage. The errors cited here dwarf the researchers mathematical masturbation, at least if the goal is relevance to the real world.
First, consider GPS ridden under tree cover, or in a canyon. The GPS signal can get spotty coverage, leading to gross errors of at least 10%. This is as real-world as it gets, and these are huge errors and I have observed them firsthand.
Consider riding a 10% to 15% grade (or even 18% for miles)—consumer GPS cannot gauge altitude with any accuracy, so it will report a shorter distance than actually traveled (because the travel is on the hypotenuse, not the shorter “leg”).
Or consider a course consisting of frequent undulating dips in the road, rising and falling 5 to 50 feet: GPS will measure that as straight-line distance (especially downhill where the descent lasts only a few second). So the distance will again be grossly in error, with greater error the steeper the slope.
Consider a step trail consisting of tight switchbacks: what are the chance that switchbacks of, say, 20-50 feet back and forth will be measured properly (and also the altitude factor, above). Particularly if the sampling frequency is too low for the time taken to negotiate the switchbacks. When hiking or bouldering where the route is constantly wavering and wandering constantly, the same sort of thing applies.
Someone should strap a GPS to Alex Honnold and see how much distance it is from the bottom of El Capitan to the top!