Thinking it over, it's also a matter of extrapolation. From practice, you can effectively fit a curve to the behavior, but you don't learn what happens outside the domain where that curve fits - and so, when you stall the wing or lose grip on the rear tires, your reactions will be exactly wrong, because you're playing by rules that don't apply any more. And yes, you can learn to fit the point of switchover and learn to fit the behavior in the new regime, in time ... but if you crash, first, it will be very expensive.