Replying to a Message

Palisade

Forum Navigation:

login

FORUMS > Off Topic

Replying to Topic: Stat arb and horse racing in Hong Kong - 24%ROI
Created On Mon Oct 28, 02 06:07 PM by SixSigma

SixSigma
Member

Posts: 64
Joined: Sep 2002

Mon Oct 28, 02 06:07 PM

User is offline

Forum:
Odds of horses winning in Hong-Kong horse races are entirely based on the public's opinions, i.e. large bets on a horse result in low odds. Teams track data on 130 variables about the horses to predict odds and identify large speads. Either teams of betters place bets or computer generated bets are placed just before the race begins.

Mathematical models are generating about 24% ROI.

Wired had a good article on it back in March.

Wired article on "Hacking the Racetrack"

Perhaps we should start our own group. Interested?

-------------------------
I'm right 19 times out of 20

matthewcroberts
Senior Member

Posts: 295
Joined: Oct 2001

Mon Oct 28, 02 08:21 PM

User is offline

I'm right 19 times out of 20

So where are the other 4 sigma?

Paul
Senior Member

Posts: 3803
Joined: Jul 2001

Tue Oct 29, 02 09:41 AM

User is offline

Yes, I would be interested...purely as a scientific study, of course

P

filthy
Senior Member

Posts: 338
Joined: Apr 2002

Tue Oct 29, 02 02:41 PM

User is offline

i would be interested. but i would also be interested in knowing exactly how you were
going to approach the problem. i don't think a purely technical approach would be
adequate. i have a nasty feeling you would need to know something about horses.

cekpet
Senior Member

Posts: 245
Joined: Jul 2002

Tue Oct 29, 02 09:27 PM

User is offline

Looks more like fundamental analysis to me rather than statarb if you insist on a market analogy.
Static arb in betting is arbing between the odds quoted by different bookies.
The more interesting variety is arbing the wrong odds of composite bets based
on the odds of the outrights.

-------------------------
You are so money...

SixSigma
Member

Posts: 64
Joined: Sep 2002

Tue Oct 29, 02 09:49 PM

User is offline

P.
I hear that the academic stuff is covered well in the following book edited by William Ziemba, alumni prof of financial modelling at the University of British Columbia. Bill Benter has a chapter about the Hong Kong betting model. The Efficiency of Racetrack Betting Models

Let's pretend for a moment that we can actually do this and satisfy ourselves that we are all so smart that we all have to wear sunglasses as we bask in our own brilliance.

The historical race data is available for posted odds(remember these are based on the bets placed), times, gate position, jockey and weight, etc. for each race for September 2001 - June 2002 at The Hong Kong Racing Journal. A back of the envelope calculation here for the simplest model of race time given the horse; 78 meets, 7-9 races per meet, 10-14 horses per race for about 6000 observations of the ~1,200 horses per season, so about 6 observations per horse - not the worst data/parameter ratio. Now we add some extra observations in the models (Benter claims ~130) are track firmness, weather, recent time trials and other related things evaluated from race tapes. These would be added as restraints in the model. Time dependence is also included in models to discount older information, so we have to add more parameters. Models are updated at the end of each season, but should be tweaked periodically. I have refined rather thin models like this by adding restraints in Non-linear Least Squares or by refining using Bayesian techniques.

The model is used to simulate the race repeatedly(probably MC) to generate the predicted odds of being in a particular finish position. Differences between predicted odds and posted odds are now exploited. All bets are charged a commission. The structured people can sort out the rather exotic bets that are available, the largest "Triple Trio" bet paid out $18MM for a team that covered 900,000 possibilities using $1.2MM in bets.

Bets are placed electronically, but should be in a way that does not dramatically alter the odds or leave footprints.

Estimates are about $US1MM to set up the databases and data feeds.

... and we don't care if anyone has a PhD or and MBA if they can do the job.

-------------------------
I'm right 19 times out of 20

Paul
Senior Member

Posts: 3803
Joined: Jul 2001

Tue Oct 29, 02 10:09 PM

User is offline

Bill Ziemba (a.k.a. "Dr Z") is one of the regular contributors to our magazine.

Can't you just tell us where to bet and on what?!

P

gjlipman
Senior Member

Posts: 1102
Joined: May 2002

Tue Oct 29, 02 10:57 PM

User is offline

I'm very interested in this (as an Australian, it is my national duty to be experienced in the art of betting on horse racing - particularly at this time of year*).

Two questions:

In Hong Kong, is the betting done on a fixed odds system (ie you are given the odds in advance) or on a totaliser basis (ie the odds which you get change as other people bet)?

Secondly, how large is the effect of the bookmakers spread? Would it be worth operating as a part of a bookmaker's team (or as a division in the TAB) so that you could reduce the spread?

(* Melbourne is currently in the middle of spring racing carnival, which will reach a peak with next Tuesday's Melbourne Cup public holiday).

SixSigma
Member

Posts: 64
Joined: Sep 2002

Wed Oct 30, 02 07:43 AM

User is offline

gjlipman:

Q1: the odds are updated on the totaliser basis. Hong Kong horse races are a pari-mutuel system. In this system, the house has no edge and no interest in who wins, essentially a zero sum game, except the track earns a 19% commission on each bet. The payouts are based only on the bets placed: there are no bookmakers. As cekpet pointed out, this is not really a stab arb problem, but a "fundamental" analysis to identify opportunities based on differences between the odds based on a fundamental analysis and those generated by people betting based on their favourite name (HYDROGEN POWER or BULLISH, really) or whatever. The constraint is that if 100,000 people bet on KNACKER and one person bets on PHARLAP*, the odds on PHARLAP will be very slim, and you stand to win if you place a bet and PHARLAP wins. You now become the bookmaker.

Q2: I'm not sure what a bookmakers' spread is.

* The only Australian horse I could think of

-------------------------
I'm right 19 times out of 20

James
Senior Member

Posts: 1559
Joined: Jan 2002

Wed Oct 30, 02 11:46 AM

User is offline

I recognize middle office, back office, and counterparty risk. And overhead. Bon chance. - James

-------------------------
"The good Christian should beware of mathematicians and all those who make empty prophecies. The danger already exists that mathematicians have made a covenant with the devil to darken the spirit and confine man in the bonds of Hell." -- St. Augustine

DominicConnor
Senior Member

Posts: 10751
Joined: Jul 2002

Wed Oct 30, 02 12:12 PM

User is offline

Not leaving a footprint would require work.
A simple model out of O'Hara would let you guesstimate the maximum return you could get for a given
effect on prices. However, electronic betting leaves big footprints, if the target actually cares.

There would be good reason to care.
A cute trick would be a variant upon front running.
Produce a list of the source IP addresses, or accounts who consistently make money.
If this scheme works then the behaviour of certain sources will stick out a mile.

Then just bet the same as them

One also assumes tha the auhorities might well have an interest in a consistent or big winner, since they want
to safeguard against less legitimate means for increasing returns. Thus they may well have such tracking already.

Actually, I'm never wrong at all, I just am right to a variable extent.

-------------------------
Discussion on the new regulations on bonuses here.

filthy
Senior Member

Posts: 338
Joined: Apr 2002

Wed Oct 30, 02 03:12 PM

User is offline

pharlap was a new zealand horse!

jimmycarter
Member

Posts: 73
Joined: Jul 2002

Thu Nov 07, 02 06:44 PM

User is offline

Except the official Jockey Club, there are a lot of underground betting houses in Hong Kong run by the trial gangs. They offer better odds, but they are illegal. I think some of those computer teams may make their bet there.

superbettingsystem
Junior Member

Posts: 4
Joined: May 2003

Fri May 23, 03 08:56 AM

User is offline

I'm Jared from Hong Kong. Since I've read a Chinese book called "super horse spirit" written by a chinese gambler whose "teacher" is Robert Moore, I put more than 5 hours a day in studying horse racing. I deeply believe the methods that the writer mentioned must work! Let's share opinions on this topic!

-------------------------
Jared

FDAXHunter
Senior Member

Posts: 3875
Joined: Nov 2002

Fri May 23, 03 08:57 AM

User is offline

This thread has been dead for 7 months.... you realize that, yeah?

-------------------------
Chiral3, kr, Nonius, Newton and FDAXHunter on a little boat drifting aimlessly on Hell's vast lake of burning sulphur.

Nonius
Senior Member

Posts: 5131
Joined: Jan 2003

Fri May 23, 03 10:58 AM

User is offline

I've developed a betting system for slot machines...

-------------------------
Nonius

I'm going to Hell with FDAX as my Kharnak.

FDAXHunter
Senior Member

Posts: 3875
Joined: Nov 2002

Fri May 23, 03 11:09 AM

User is offline

I've also developed a betting system for slot machines... you need a crow-bar and a screwdriver to make it work, though.

-------------------------
Chiral3, kr, Nonius, Newton and FDAXHunter on a little boat drifting aimlessly on Hell's vast lake of burning sulphur.

superbettingsystem
Junior Member

Posts: 4
Joined: May 2003

Mon May 26, 03 12:13 PM

User is offline

Hi all! I deeply believe horse racing is "mathematics". For example, a coin, if someone asks you to choose head or tail and the odds of head is 1 and tail is 1. Then you don't have any advantages. However, if the odds of head is 0.5 and tail is 4. We know that the probability of tail is 50%. If you bet tail in long term, you must be a winner. It seems that it is impossible in real life. However, horse racing is a kind of this game. For exdample, if the winning probability of horse A is 10%, the reasonable odds of horse A should be 10-1. However, in reality, the odds of horse A is only 3-1. This horse is so-called "false favourite". If we can find out which horse is underbet at the same race, we will be the winner in long term. The problem is how to calculate the probability of a horse. How to calculate in a right method. That's why horse racing is "mathematics". I am really interested in this topic.

-------------------------
Jared

kapital
Senior Member

Posts: 236
Joined: Jul 2002

Mon May 26, 03 06:26 PM

User is offline

superbettingsystem,

don't know HK, but in the U.S. winning at the track is easy. but you should know it sucks - no fun and not a lot of money.

go to the same track every day. over time find out what the owners & trainers want to do with each horse. after some time (depends on the size of the track) you will know 7/8 of the horses in a given race. only use the stats in the racing forum to quantify what you know qualitatively about each past race and how the horses compete with each other.

the way to win is to narrow the list down to 1 or 2 horses that really have a chance to win. if you're not talking the kentuky derby, there are usually a few horses that aren't really in the race to win (maybe the trainer just wants to see how they run in a new class or maybe you know one of the horses is still a little gimpy from a collision 3 weeks ago and ran too hard in the last race, whatever). of the rest, just use what you know about their previous showings and what the owner/trainer's intentions are. some of this information (training injuries, 1 phrase describing the last showing, etc) is in the racing form, but there's no detail and it only goes back a few races. if you've seen the races and you know the horses, you will be able to size up the chances better than 60% of the other people at the track who are just going off the forum (or the free track sheet) and the other 20-30% who are just betting on names or whatever.

bet 1 of your two horses win-place-show for $100 ($300 bet) and over time you will make enough money to live and to keep playing. you can't really talk ROI terms because it won't scale, but you can make about $50k a year this way.

just make sure not to start betting money until you've really done your homework (not statistics). good luck to a wonderful career in scalping suckers (uh, or people with a real life) at the track!

superbettingsystem
Junior Member

Posts: 4
Joined: May 2003

Thu May 29, 03 09:42 AM

User is offline

Hi Kapital,

There are only two race courses and only 1200 horses in Hong Kong. It seems easy to collect datas and that's why there are so many international gamblers in Hong Kong. If you have time, take a look with this website www.hongkongjockeyclub.com/english/index.asp
By the way, how big is the pool in each race in the US? You know, in Hong Kong, about 100,000,000 Hong Kong dollars bet in each race. So crazy!

-------------------------
Jared

FORUMS > Off Topic

Forum Navigation:

© All material, including contents and design, copyright Wilmott Electronic Media Limited - FuseTalk 4.01 © 1999-2010 FuseTalk Inc.