I see a lot of people talking in generalizations and estimates about these figures, and that really got me wondering. Is there a place where I can find such numbers? Are they even available?
And part 2... if there isn't such a place to get those datasets... would anyone here mind posting some up?
There would need to be a standardized way of measuring, though. Those scanguages just might do it. There would need to be a standardized method (and format) as well. Once some samples are taken in a standardized way, it would be a matter of a little calculus and all that guesswork could be replaced with valid, reliable data (and corresponding 'rules') on a per-model basis. But that's if it already doesn't exist.