I agree. In theory it seems studying the case that deviates more from the average (in this case, great success) should yield more information. If most widgets are small and a few are big, and you want to know what properties of widgets correlate with size, and you can study only one widget, you should study a big widget.
I agree. In theory it seems studying the case that deviates more from the average (in this case, great success) should yield more information. If most widgets are small and a few are big, and you want to know what properties of widgets correlate with size, and you can study only one widget, you should study a big widget.