Some parts are interesting, and the writing can be entertaining, but the book is short on insight and clarity and long on tedious tables and uninterpreted computations.
Buy this if you already know probability and would like to see -some- applications and cute games.
Don't buy it if you want insight into particular games; especially, the blackjack and bridge sections (and meager poker section) have virtually no value.
I am a graduate student in mathematics, and enjoy probability theory and games: I should be the ideal audience.
The math is no problem for me, but much is boring, and much time is spent writing huge tables without giving much insight.
Research articles in statistics are easier to read, and far more informative.
The math background is awful: if you don't already know it, don't learn it here.
[Instead, see "The Cartoon Guide to Statistics", or Feller's "Intro to Probability"]
The writing is willfully obscure and florid (though, admittedly,
entertaining): gymkhana, panjandrum, kubiagenesis?
My main objection is the lack of insight: the author does (mostly) correct computations and statements but seldom shows much depth of understanding and rarely conveys any to the reader.
Rather than answering questions or giving examples that convey the meaning of the theory, how it lets you understand questions, Epstein does many unillustrative examples.
This book won't teach you to understand games and gambling, which it could do, and should do.
At best, it provides a basis from which you can (after too much work) begin to understand games. This is not because the subject is that hard (at least not what Epstein covers) -- it's because the material is undigested and Epstein is a poor expositor.
If you want to get something out of this book, be prepared to do the work that Epstein hasn't, and to look at more modern and insightful references.
Here's an example: how many times do you need to shuffle a deck before it's essentially random? Very natural question, of big interest in gambling. Epstein gives a very slick argument, one of the gems of the book (measure entropy of a shuffle) that you need at least 5 shuffles -- but beyond that just writes some equations for 2 shuffles of a 4-card deck and says that a computer would help, and instead tabulates that 18 perfect shuffles of a 58-card deck return it to the original state.
The rest of the book is like this: some question begging for study, perhaps an insight, and then irrelevant and pedantic computations and tables.
There are gems in here (it's a grab-bag), and the writing is often amusing, but it's a frustrating read: it could be so much better.
