So presumably if the naive interpretation is wrong, then the bias should be predictable. Any ideas how to go about determining which permutations are more like

So presumably if the naive interpretation is wrong, then the bias should be predictable. Any ideas how to go about determining which permutations are more likely than they ought to be?