Kelly for Horse Racing

**LT Profits** · 05-11-11, 09:59 PM

Kelly is impossible to apply in parimutuel betting because you never know exactly what the closing odds will be.

But pertaining to your post, you are right that any bet makes no sense when model odds are greater than the offered odds.

**CoachB** · 05-12-11, 02:56 AM

LT,

Thanks for your imput.

To say it is "impossible" apply Kelly is a bit harsh.

The model can be initially set up using approximates, and in fact fixed price betting is available in the markets I am considering, so the final numbers can be locked in well before post time if required.

CB

**LT Profits** · 05-12-11, 03:09 AM

Originally posted by CoachB

LT,

Thanks for your imput.

To say it is "impossible" apply Kelly is a bit harsh.

The model can be initially set up using approximates, and in fact fixed price betting is available in the markets I am considering, so the final numbers can be locked in well before post time if required.

CB

Gotcha, then refer to my second sentence.

I do stand by my first sentence, especially at tracks with small pools, but if you have fixed betting available, then parimutuel is moot.

**matekus** · 05-12-11, 08:05 AM

Taking Chances

CB,

See Haigh J (2003), "Taking Chances" [ISBN-13: 978-0198526636] Appendix V (Kelly Strategy) for a worked example using horse-racing.

matekus

**ByeShea** · 05-12-11, 08:11 AM

No one can beat horseracing with its ridiculous takeout. No one.

**smoke a bowl** · 05-12-11, 12:27 PM

Originally posted by ByeShea

No one can beat horseracing with its ridiculous takeout. No one.

There are exchanges such as Matchbook, Betmaker,Betfair, that offer yes/no prices on all horses and those markets are very beatable if you are better than your competition.

**u21c3f6** · 05-12-11, 12:53 PM

Originally posted by CoachB

...
Given the odds available are $3.11 which are < the required $4.35 I would have thought Kelly would say no bet due to negative expectation.

Yet in the final model the author has calculated a bet for this horse of 4.30% of bankroll.

Can anyone shed some light on this, have I missed something or does this paper contain an error. If it is the later what changes are needed to fix the model?

Cheers

CB

If you were to wager on #10 only, then you are correct, there would be no Kelly wager. However, hedging (dutching) is quite different. There are many times that it is actually beneficial for your bankroll to hedge, even with –EV wagers. I assume the paper has the math correct but I did not check it. I will give you an example with “easier” numbers to try to clarify the concept.

Assume there is an event which has three outcomes, A, B and C. The odds are as follows:
A +300
B +300
C -105

Slightly –EV if you try to dutch all 3 outcomes. Now, let’s say you “know” that the real probabilities are as follows:

A: 30%
B: 20%
C: 50%

Of the three outcomes, A is +EV while B and C are –EV. How should we wager? Applying Kelly we see that we have a 20% edge if we wager on A. 20%/3 (edge/odds) = a Kelly wager of 6.67%. After 100 wagers assuming that we hit exactly 30 winners, a $1,000 bankroll would become $1,896.76.

Now B is 20% -EV and C is 2.5% -EV. If we want to see if we can improve on the above wager, obviously we want to look at adding C and not B from both a win % and the fact that C has a much smaller –EV. There is of course a calculation to derive the exact Kelly numbers but again I will set my example up for the purpose of clarifying the concept and not a claim that this set-up is optimal.

My new wager would be 1.00 to win on A and 1.05 to win on C. This means that there will be 30 trials where I would win 1.95 (3-1.05), 20 trials where I would lose 2.05 and 50 trials where I would break even. This means that we have a 60% (30 out of 50) chance of winning 1.95 and a 40% chance of losing 2.05 not counting the 50 breakeven wagers. This amounts to 60% wins @ approx. -105.13 or a Kelly wager of approx. 17.95%. This new wager applied to 100 wagers of which 50% are breakeven and assuming we hit exactly 30 of the remaining 50 would result in a bankroll of $2,164.86 or a much higher growth rate than just making the +EV wager.

Joe.

**CoachB** · 05-12-11, 11:36 PM

Originally posted by matekus

CB,

See Haigh J (2003), "Taking Chances" [ISBN-13: 978-0198526636] Appendix V (Kelly Strategy) for a worked example using horse-racing.

matekus

Thanks for the heads up.

Found the example on Amazon.

Not written in a user friendly way* so I am still working my way through it.

*Academics can't help themselves!

Cheers

CB

**CoachB** · 05-12-11, 11:37 PM

Joe,

Still working through your note too. Thx

CB