An edge point's probability of being hit should be proportional to the length of every path leading to that edge point. An area closer to many short black paths and many long white paths will show black expansion (and vice-versa). So I suspect that any variation of the central line from a straight bisection of the circle should get hammered out over time.