Obviously not the point of the post, but this doesn't seem like a very reasonable thing to calculate, statistically speaking.

Also, I wonder if there is some way to use branch-and-bound to look at fewer combinations.