I missed this question.
He did a great job of making it interesting. It bugs me because
his evidence that his system 'works' is quite weak.
One can see that he is a computer scientist rather than a statistician.
I haven't used his book at all. I (and many others) were using
a monte Carlo simulation long before. I think his method for estimating
point label probabilities from game level results is quite a bit off.
