You'd think, but I'm currently looking at a 161/863 split which has been pretty stable for a while now. I thought at first that ball speed is being scaled by te

You'd think, but I'm currently looking at a 161/863 split which has been pretty stable for a while now. I thought at first that ball speed is being scaled by territory size to cancel out the negative feedback, and the "winning" ball is moving noticeably faster than the losing one, but I can't see that being deliberate in the code. (There is some randomization of speed on a bounce, but it ought to be pure noise.) EDIT: 120/904 now.