Deterministic Schelling sociology

I'm utterly unfamiliar with the model proper and all more recent work exploring Schelling's segregation models besides what David [1] patiently explained to me. But I find myself strangely, repeatedly, clicking-to-randomize his Schelling Applet [2] and trying, by observing the random dot scatters on each new "blank" canvas, to pick a pixel that will be green after the system emerges into its steady state.

With just a little bit of guessing at visual clustering, I can guess the right answer more times than not, which probably isn't surprising to anybody who knows the theory and its implications. But then that makes me want to further pre-cluster and aggregate little cluster nodes and bigger ones to study before and after connectedness levels and see if that can help predict anything about social graphs and data webs.

If, perhaps, I had any idea how to do that, that is.