Oh yeah, I've thought about it. This is why I've been talking about systematic vs. random error for a long time now. We have different ways to deal with different types and sources of error, and it would be a mistake to treat systematic sources of error as if they were random. For example, most people use the classic regression method as if errors met the Gauss-Markov conditions; but systematic errors obviously don't. The standard way experimentalists try to avoid systematic error is to control the experiment more tightly, like testing on a dead calm zero wind day. But a second way is to measure and then do statistical controls post hoc. This is where air speed and yaw sensors come in, but other sensors do, too.

In the meantime, even if you have a wind speed and direction sensor, there may be other systematic sources of error that you will want to identify. VE appears to work in that context too. Here's something I've long wished GC's Aerolab would do: I wish you could overlay laps, or mirror reflect out-and-backs and then overlay them. That would make it easier to spot a passing car or a bobble. Then I wish it would make cutting-and-pasting to excise obvious passing cars or bobbles. Those are "non-random" (in a statistical sense) so you want to skip over them instead of treating them like a source of random error. That would be a lot handier than tossing the entire trial run. There are other things I think would be helpful, but I'm not a C++ programmer so I've never been able to do them myself.