Proposing a Profitable and Useful Horserace Betting Gizmo

[Borodog]

David,

Not being one to play the horses, the "Place" and "Show" probabilities and odds in the above tables are incorrect. What I've labelled as "to place" is the probability that the horse will take exactly 2nd, not 2nd or 1st, as it should be. Similarly, the "to show" probabilities are also incorrect, showing the probability to take exactly 3rd, rather than 1st, 2nd, or 3rd. The odds are similarly defective. I will produce a corrected version. My apologies.

Corrected tables:

Probabilities

Horse 1 to win: 0.3043955
to place: 0.5781337
to show: 0.7780613

Horse 2 to win: 0.2730526
to place: 0.4802549
to show: 0.6534196

Horse 3 to win: 0.1187247
to place: 0.3465173
to show: 0.6226993

Horse 4 to win: 0.0845282
to place: 0.2554323
to show: 0.4981509

Horse 5 to win: 0.2192990
to place: 0.3396618
to show: 0.4476689

Exactas
Win\Place 1 2 3 4 5
1 ------------- 0.0947495 0.0930645 0.0663041 0.0502774
2 0.1126503 ------------- 0.0677799 0.0481196 0.0445028
3 0.0474221 0.0306293 ------------- 0.0253095 0.0153638
4 0.0312456 0.0201805 0.0228833 ------------- 0.0102188
5 0.0824202 0.0616430 0.0440649 0.0311709 -------------

Odds

Horse 1 to win: 2.2851996
to place: 0.7297037
to show: 0.28524578

Horse 2 to win: 2.662298
to place: 1.0822276
to show: 0.5304102

Horse 3 to win: 7.4228473
to place: 1.8858588
to show: 0.60591155

Horse 4 to win: 10.830372
to place: 2.9149318
to show: 1.0074239

Horse 5 to win: 3.5599842
to place: 1.944105
to show: 1.2337937

Exactas
W\P 1 2 3 4 5
1 ------------- 9.5541460 9.7452350 14.082024 18.889652
2 7.8770290 ------------- 13.753636 19.781553 21.470497
3 20.087215 31.648478 ------------- 38.510857 64.088066
4 31.004507 48.552788 42.699990 ------------- 96.858850
5 11.132948 15.222442 21.693800 31.081203 -------------

[mindflayer]

Wow, thats some fast footwork Borodog.
Do you work for pokertraker or something?!?

[mindflayer]

Yes you missed the all critical factor...
(the one i make my informed betting decision on)

7. Did the horse take a crap on the walkup to the gate.
The bigger the crap, the bigger the bet!

[Borodog]

[ QUOTE ]
Wow, thats some fast footwork Borodog.
Do you work for pokertraker or something?!?

[/ QUOTE ]

My background is in complex numerical simulation.

[Borodog]

This has really turned out to be a pretty interesting topic. I added a few horses to the race, to total 8.

Lucky 100. 0.41
to win: 4.267885
to place: 1.7528788
to show: 1.018141

Seabiscuit 100.1 0.55
to win: 4.2630887
to place: 2.0774524
to show: 1.3762829

Dog Food 100.2 0.32
to win: 17.459414
to place: 5.9299493
to show: 2.9328969

Glue Factory 100.3 0.35
to win: 24.859638
to place: 8.689058
to show: 4.425933

Trigger 100.4 0.85
to win: 4.8806233
to place: 2.9339876
to show: 2.2259896

Sugar Foot 100.5 1.
to win: 4.529377
to place: 2.9429455
to show: 2.3296108

Artless Dodger 100. 0.25
to win: 7.7726145
to place: 2.3396676
to show: 1.0264102

Hoofed Victory 99.98 0.1
to win: 15.017555
to place: 2.8903558
to show: 0.84739655

Lucky, Seabiscuit, Dog Food, Glue Factory and Trigger are David's original 5 horses. Notice that Trigger and Sugar Foot, the two horses with the worst average times, have the 3rd and 4th lowest odds against to win, because of large standard deviations. And the fastest horse in the field, Hoofed Victory, is 15-1 to win, due to a small standard deviation. Compare the Artless Dodger to Lucky; they have identical average times, but the Dodger's consistency hurts him significantly to win. However, the vulnerability of a faster horse with a small SD to win doesn't translate into the odds against showing.

Does anyone have any actual data? I can now run an arbitrary number of horses if I have their average times and standard deviations. Give me a field and the lines, and I will pick out the profitable bets for you.

[mojobluesman]

One of the issues is that you need to define "average time".

The problem here is that many horses have fluctuating ability and not just fluctuating times. You need to seperate the two and that's no easy task.

Some of them are younger and improving.

Some of them are older and declining.

Some either have minor injuries that are causing decline or are recovering from one and getting back into shape.

Some have had trainer changes that improve or worsen the horse.

Also, there are track condition, surface, distance, and other issues that contribute to standard deviation that might not actually be reflective of SD under today's conditions.

In practice, most expert handicappers use subjective analysis to try to determine which races are representative, but there will be disagreements.

[Superfluous Man]

I don't know anything about horse racing, but I do know that the existence of such a gizmo was the premise of the Joseph Heller short story "Bookies, Beware!"

[BlindMansBluff]

David,

I sent your interesting posts to a few of my friends in the business. Here's a response from Nick Mordin, printed with his expressed permission.

-David McKenzie

Dear David,

The first problem with this idea is that it makes the common but fatal mistake of assuming that racehorses are numbers on a page. In fact they are flesh and blood creatures with different physical and mental characteristics. They don't deviate from their own standards for identical reasons. They deviate for reasons that are unique to themselves.

For example there is a horse running in the Hong Kong Cup tomorrow called Pride. She's not very big and hates crowding in a race because she comes off worse in traffic. She also needs a run within 30 days to get her fit. So far she has run seven times in fields of nine or less when she's had a run within thirty days. She won six of these seven times and finished a neck second in the Arc in her sole loss. Tomorrow she races in a field of twelve and she's lost all six times she's run in fields of ten or more in Group 1 or Grade 1 company. In addition her last race was back in October, way more than 30 days ago.

You would not gain this insight into Pride without reading what her trainer and jockey had to say, knowing her physique and watching her races. In other words you would not be able to predict her standard deviation without carrying out a good deal of handicapping.

The second problem with this method is that it assumes each horses chances can and should be expressed in percentage terms. I don't think they can. In my experience there are simply far too many variables. The best we can do is look out for horses like Pride and bet against them when they are heavy favourites like she is for the HK race. Then we'll know by inference that the odds available about other horses in the race are going to be bigger than they should be.

If you analyse the bets made by any professional gambler you'll find that, while some of them may make a 'morning line' and try to predict the odds for every runner, all of them, without exception, make their heaviest bets in races where their handicapping tells them that a short priced favourite is likely to lose. Bookmakers work the same way. They trade on the percentages most of the time but make most of their money when they 'take an opinion' about a particular favourite and effectively bet that it will lose by holding its odds in the face of large volumes of bets.

No-one can make an accurate odds line because no-one can accurately handicap every horse. What you've got to do in order to make money is to oppose horses that you can handicap when they run in unfavourable circumstances at short odds.

The average professional gambler or bookmaker probably only understands about 1% of the horses they deal with well enough to figure out reliably whether they're going to win or lose. They hold their money till one of these horses comes along and bet accordingly.

Regards
Nicholas

[Borodog]

[ QUOTE ]
One of the issues is that you need to define "average time".

The problem here is that many horses have fluctuating ability and not just fluctuating times. You need to seperate the two and that's no easy task.

Some of them are younger and improving.

Some of them are older and declining.

Some either have minor injuries that are causing decline or are recovering from one and getting back into shape.

Some have had trainer changes that improve or worsen the horse.

Also, there are track condition, surface, distance, and other issues that contribute to standard deviation that might not actually be reflective of SD under today's conditions.

In practice, most expert handicappers use subjective analysis to try to determine which races are representative, but there will be disagreements.

[/ QUOTE ]

Luckily, none of that is my problem. I don't know dick about horses. I imagine it would be extremely difficult to construct the data. However, a knowledgeable handicapper should be able to do better than no data at all. Hence, for a knowledgeable handicapper, this technique could only possibly help. The trick of course is: Is the handicapper actually knowledgeable enough to produce inputs that are in fact better than no data at all, or does he only think he is?

[mojobluesman]

Your friend knows what he's talking about.

I would simply add, sometimes it's possible to know that a particular horse is better than he looks on paper just as we can know that a favorite is highly vulnerable. The horse that is better than he looks on paper may not even be the most likely winner, but he can offer good long term value.