I too would be interested to see how he arrived at the .55 figure
for post 8. After seeing your post, I did a quick mod to generate
my own numbers.
What I got is approx .49. However I then tried a very reasonable 3% bias for the receiver, and that did work out to about .55.
It occurred to me that if you wanted to bump up the sophistication a bit you could have a massive table for each trifecta number, eg if # = 126, post 6 show figures might be 2.30 and 1.26. You would probably get very slight improvements in accuracy that would make almost no difference in performance, tho.
Still, I think I'll order his book just for curiousity's sake.