DiscoGary wrote:
hunt the big game wrote:
The GPS measurements were amazingly accurate and consistent. They differed in length based on where the runners ran on the course primarily due to where they were in the pack. You can view measurements from the day prior also for runners when they were able to run closer to the line.
How do you know they were accurate? You are making the case that they were precise.
I think we need to discuss GPS accuracy versus precision because it looks like GPS precision has improved in the last 15 years and a lot of people are claiming this means the accuracy of GPS measurements is getting better. This is going to be hard to swallow, but depending on the number of turns, the accuracy might be getting worse. In fact, <b>a perfectly precise GPS unit will guarantee an inaccurate measurement on courses with turns!</b> I've tried to avoid this level of detail because without diagrams its hard to explain what's happening, but here goes.
Let's start with 3 basic concepts for GPS measurement.
1. GPS takes location measurements at a specific sampling rate. In many GPS units that sampling rate can be adjusted.
2. GPS measurements have an inherent location uncertainty which introduces location errors for each measurement in a 360 degree circle surrounding the actual location. So the actual location maybe to the side, front, or back of the reported location.
3. GPS calculates the linear distance between two measurements, so error is introduced on turns due to the fact that the turn is truncated. Maybe there are GPS units with curve fitting algorithms, but I am not aware of this. If not, they should be doing it.
So let's do a thought experiment with our perfectly precise GPS unit. On the straights it doesn't matter how fast or slow you go as long as you stay near the SPR your measurement will perfect. So far so good.
Now we get to the turn. The faster we go around the turn the fewer number of sampling points we get and the more truncation will occur, guaranteeing that the course will measure short ... assuming a perfect GPS unit. The only way to get very close to the actual length is to speed up the sampling rate or slow waaaaay down around the turns so that the sampling points are almost on top of each other.
I know what you're thinking. "Thanks Mr. Disco. Will do. I'll be right back." .... Wait!
The problem is that GPS unit are NOT perfectly precise. They do have an inherent location error, and that error can accumulate with each sample so the slower you go the more error will accumulate, unless they are curve fitting, and they might be, but then the curve fitting needs to be checked. So how do you verify the unit is curve fitting?
Glad you asked. You set the unit to its highest sampling rate and go very slowly on a straight path for as long as you can tolerate. Lateral measurement errors should be piling up without curve fitting, making the path seem longer. Then go as fast as you can on the same path. If both measurements come out to within a meter, then the unit must be curve fitting, which means its throwing out distortions caused by lateral measuring errors.
In general a purist won't want to hear that our measuring tool is actively changing measurements along the way to make things come out right, but I'm willing to keep an open mind on this because a very precise GPS unit with a properly calibrated curve fitting algorithm would definitely be "good enough for cross country", and might even be good enough for keeping records. But I gotta see the proofs first, and that means testing the theory against the current gold standard, the tape.
At this point, a whole bunch of the GPS measuring guys should be pulling out that tiny manual that came with it trying to find out if their unit does curve fitting. Let us know.
You also might be thinking that I'm back tracking from the tape measuring push. A little yeah. Because until I started to explain this it hadn't occurred to me that GPS units might be actively trying to remove measuring errors. The more I think about it, the more I think they probably are because the people building these things have to have thought about this, so I really want to see this put to the test.