"Forecasting Methods for Horseracing"

[Rick]

Has anyone read this book by Peter May? It's available at Amazon and Gambler's Book Shop now and sounds pretty interesting.

[GameTheory]

I looked it up at both places but can't find a description of it either place (the page at Gambler's Book gave an error).

[Rick]

GT,

It's in "New Arrivals" at GBS. I think it's the same price at both places.

[GameTheory]

Aha...

Sounds like a book I'd write, if I were to write one. I've got a stack of material literally 3 feet high on such techniques. I don't need to add to it at the moment.

If you get a copy though, let us know how it goes.


Here, I found a link to his home page:

http://www.pjmr.freeserve.co.uk/

He's got a couple of papers in the neural network section...

[Rick]

GT,

Thanks for the link. I wish I had your 3 foot stack. Mine is considerably smaller. What do you consider to be the best of them?

[Lefty]

With the word Forecast in title I figured it was for British Racing. Looks like I was right.

[GameTheory]

Techniques? I think neural networks are the least promising actually. At least in terms of the amount of work you have to put into them to get a decent result. But they may work great for others.

One of the quickest and easiest AI type methods that produces good results are automatically generated "decision trees". Mathematically, they are pretty simple, most of the time using information theory/entropy (remember that formula from your buddy Epstein?) as a basis to pick useful features. Here are some keywords to search on: "decision trees" "C4.5" "information gain"

[Rick]

GT,

What do you think of Epstein's entropy theory? Many years ago I tested it at Santa Anita and horses at odds of 8-1 showed a profit over a small sample. I used a simplified method where I played races with 9 or more entries and no horse at less than 3-2, which pretty much guarantees you a low entropy.

Lefty,

That's the only thing I'm worried about. How relevant will the book to US racing when it was written by a British author. But on the other hand, the best paper on horse racing models that I've ever read was written about Hong Kong racing.

[GameTheory]

I don't anything about Epstein or his theory. I don't have his book, although I'd like to read it. All I know is you once gave me the entropy formula, which is a standard formula in information theory.

[Rick]

GT,

Yeah, I don't know how useful it would be, although if someone is into betting longshots it might be worthwhile.

[Jake]

I have the book. I ordered it several years ago directly from England, because I thought it might be worth the effort. It turned out to be an expansion of a thesis paper, and one done at a very simple level. It covered some very elementary neural net stuff with a very simple factor set, and overall was pretty disappointing. I wouldn't recommend it to anyone, especially anyone wishing to use neural nets.

On the other hand, a few months back I ordered a book by Nick Modrin called Winning Without Thinking, which deals with racing systems. It covers both English and US racing, and is one of the most interesting books I have seen on racing. It was discussed several months ago here on the board, and I think was even recommended by a poster from Hong Kong. Expensive to have sent over, but just an excellent read.

[hdcper]

Hi Jake,

Haven't heard from you in awhile and John said he emailed you a few days ago and it came back. If you see this, could you shoot me an email with you new email address?

Hope all is well with you and yours,

Bill

[Jake]

Originally posted by hdcper
Hi Jake,

Haven't heard from you in awhile and John said he emailed you a few days ago and it came back. If you see this, could you shoot me an email with you new email address?

Hope all is well with you and yours,

Bill


It's still the same. It was down for a day while the mail server was being switched over, but it appears to be working fine since then...surprised to hear I was missing an email.