To my surprise I could not really detect much correlation between mobility and sub-tree sizes. (Tested by trying if the sub-tree size divided by some power of the mobility would have smaller variance as the raw sub-tree sizes.) And if the sub-tree sizes do not correlate with the mobility of their parent, why would they correlate with each other? Perhaps they correlate mostly with the mobility of the grand-parent. Or perhaps with the opponent mobility in the parent.
![Idea :idea:](./images/smilies/icon_idea.gif)