UK Horse Racing

Book Review - Calculated Bets

"Computers, Gambling, Mathematical Modeling To Win" by Steven Skiena (Cambridge University Press)

Intoduction to Jai Alai

When one discovers a mathematical text book on the subject of betting, and from an author who has claimed to have made a profit from betting then, well, one just has to sit up and take note; or at least read the book.

Calculated Bets is the story of how Skiena has written a mathematical model to predict games of Jai Alai. Jail Alai is that fascinating sport, which I have until now have only associated with Miama, is played as either singles or doubles on a large court, a fronton, and wherein the players use baskets tied to their arms to hurl the ball down the court rather than using the more normal racquet.

The book goes into game's history briefly, where it actually derives from the Basque region of Spain, and the rather strange scoring method used in the game.

The history and the background of the sport takes the first portion of the book and, to be honest, is the clearest and most interesting section.

Whodunnit? - a mystery

The reason that I got hold of this book was that I was curious to see how he processed his data to make his selections. How did he make his selections and methods and how to they compare to ours?

The book fell apart about here. A few basic half-hearted bits on statistics were presented. For example, he touched on how to simple set a curve to set of data-points. This could have taken a book in its own right but although Skiena mentioned that one could do this there was no indication that he used data-fitting in his program and, if so, how he managed to fit the data (which data?) to the database.

In fact, the question 'How Did He Do It?' wasn't satisfactorily answered at all. Of course we at UK Horse Racing are not going to tell you our methods in any detail; then then again we're not going to charge you over ten pounds to tell you that!

Skirting the subject

Throughout the book Skiena would touch on one of three themes; the first being very basic mathematics, or his own brand of software philosophy or how the code was written for him.

Given that Skiena is a professor at New York University I would have expected him to be able to write the simplest of routines, but, instead, he seemed to have used his students because, as he freely admits throughout the book, that he is no developer. Even though he seems to be in awe of those who are able to put a few commands together as he quite willingly rabbits on about the Obfuscated C Contests which has no place here in this book and only serves to confuse those who don't know nor understand why this needs to be done.

Neural networks are presented and then dropped again in the space of a few pages. Even the basic principles of neural nets are skipped over leaving the reader who has bever come across Neural Nets to wonder what the hell was that all about. And, his rant against Microsoft is best left forgotton because, firstly, there is no place for it here in this book and, secondly, most of what he writes in that passage is garbage. The only reason for its conclusion, I think, is to satisfy his ego and to join the ranks of those ultra-hip and trendy techies who hate Microsoft. At least some of those guys know and understand their agenda and have the technical knowledge to argue their corner; on the other hand the reader is just left to feel that Skiena is jumping on the first passing bandwagon that he can see.

Returning to the book; nowhere does he describe the inner working or the basic mathematics of his code other than he uses a Monte Carlo method to replay each match umpteen thousand times. There is simply no link between the maths (which maths?) and the model whatsoever and this was a major failing of the book.

The proof of the pudding

Skiena is supposed to be a mathematician in the Computer Science faculty of the university and surely it's not impossible that he would know of things like basic probability. Anyone who knows the merest thing about statistics, particularlyu those who indulge in any form of gaming, is that losing runs occur and, thankfully, hot winning streaks can counter.

For example in tossing a fair coin it is possible to lose three times in a row one time in eight. That doesn't mean that each eighth toss of the coin will mean a start of a three coin losing run but that it could do. This is basic stuff and something which Skiena should have understood or at least mentioned in his book. So when Skiena gives us a report when he turned on his Model, "Maven", and ran it over a few days and it made a profit.

And that's it. Other than the staking system would have scared The Man of Steel.

What did he do then? He stopped the system when it hit a losing run and then wrote the book. Even when he wasn't betting on the game he never kept a database of results and when he wrote "Maven still sends me email every morning, and of date it seems our simulated bets have been winning somewhat steadily than before".

This is no different from asking the man in the betting shop how well he does at the horses; you just know that he will say "I am about breaking even".

And that's it. No mention of the number of bets struck, the losing runs or the return on investment. No statistical breakdown of how the model performed. The book was finished off at this point in a most unsatisfactory manner.

Excuses, excuses

The excuse Skien trots out for the start of the losing run before he turned off the machine? That there was 'software rot' or that the data changed.

The former is utter garbage and, if the case of the latter, then he should have jolly well have gone to sort this out.

I note that he doesn't say that what happened to his losing run. How much did he lose? All he tells us where the betting bank reached its highest point. But, again, Skiena didn't look in the possibility of a losing run which is statistically accepted and so he just gave up and turned off the auto-dialler which talked to the bookie's computers.

So, then, does his model work? I have no idea though he does claim to have made a profit (which he gave away to charity) before writing the book.

Conclusion

Other than as an excellent summary of the game of Jai Alai this book cannot be recommended at all. Apart from a few interesting small sections in the book there is little or no information in here with which anyone could use to start off a Model of their own.

On top of everything Skiena's Model, Maven, seems to be a simple one. Nothing in there could convince me that it could be adjusted to model a more complicated sport such as racing. Of course he mentions racing and Skiena dismisses it as a possibility (too complicated for him, perhaps?) and this would only be the simpler American racing rather than the more exotic and colourful racing within the British Isles.

And, reading the book as a more than half way decent software developer who's been around the block a few times I found Skiena's tirades pointless, tiring and utterly well off the mark. What seems to be true to form for academics is that he seems not to have a grasp of the real world at all. Though one thing the book does well at; that how much it has managed to dissuade me on betting within the United States.

ISBN: 0-521-00962-6

Available from Amazon.co.uk and Amazon.com.