Does anyone has a computer model for Hongkong horse racing forecast?
Or want to build one?
I wanted to tackle handicapping this way some years ago so got this book https://www.amazon.com/Precision-Sta.../dp/1432768522Originally Posted by peterlee
The problem I ran into wasn't building a data modeling algorithm, but efficiently backtesting it. I like the game, but just a long-time loser in it, so lost my desire to play it seriously.
The computer teams getting rebates really make me feel like a sucker. They really dilute the pools in the horizontal wagers.
They have access to APIs that can look deep into the horizontal wagering pools and get access to data that normal people can't.
It is only a matter of time before these asshole greedy racetrack managers that cut these deals for more handle have lost every normal player and are left with
nothing but computer teams duking it out for .004% ROI + their rebates.
That book is translated from a Chinese version, writer is from Hongkong. It tells some of the math, but not teach you how to build a model.
Bill Benter, Zeljko Ranogajec, and many other models, started from a little yellow book that Bill does not want to tell.
https://www.youtube.com/watch?v=YOVrZrJ-wtc
Maybe I am misunderstanding your goal, but here is my 2 cents
Building a handicapping model should just be a matter of finding a handful or thousands of quantifiable variables and manipulating them by whatever statistical methods you choose to estimate a horse's true probability. If you know a programming language you can build a program that has a basic UI that can receive what you input and write some functions that crunch the data as you choose.
Caveats- I went down this path several years ago and hit a brick wall.
1. Building a program like this is utterly useless if you can't properly backtest and prove your methods are +EV long-term. If you can back test it is a rinse-and-repeat process.
Set variables and methods up and backtest. If your backtesting shows long-term losses you will tweak your methods and try again.
2. It isn't 1985 so manually entering data into a custom UI is not ideal. There are APIs you can access to automatically pull in your data but they cost a ton (at least the ones I came across years back).
3. At the end up the day the quality of your model will be tied to the math abilty of the developer.
I hear ya loud and clear, These days I just play a few times a year. TC series and breeders cup day.
Like I said yesterday I like the sport, but it has proven to be kryptonite to my bankroll over the years.
I just can't grind my BR up as I can in blackjack or poker and the occasional options trade.
I have my own model, it took me 18 monthes full time job to build it, in early 90's.
There are twenty variables in my model, while Bill Benter created over two hundred and uses about 130 variables in his model.
I once helped MIT John Chang's team to build one, but the project was terminated after two months, before anything came out.
To build a model, that needs a programmer, a math expert, a route how to build it. And where to place your bets if the model is profitable.
+++
My goal, so many experts around here, then why not build one?
Yes, I do backtest. In two ways:
1. Say I have 10 years of horse racing data. Use year 1 to year 8 for the model building, simulating bet on year 9 and 10.
2. Take 80% of the races randomly from the 10 years data to build the model, backtest the 20% remaining races.
That an amazing accompishment. I never followed through with mine. I am going to Florida for about 10 days.
When I get back I am going to dig though my office and try and find the old rough draft outline of ideas I created
that I was going to try and build my model off of. I would llke to send it you and get your feedback.
If you would like email me [email protected].
Posting Permissions
Reply With Quote
Bookmarks