Optimal bets for dog and horse races

Andrew Colin PhD, works through the mathematics for how to analyse and optimise your betting on pari-mutuel and tote pools for horse and dog racing.

Categories: Funding, Horse Racing, Prices, Professional, Statistical models

Given a set of probabilities for the runners in a multi-player competition, how much should you bet on each runner to maximise your expected profits?

The standard reference paper for this topic was published by Isaacs over 60 years ago[1]. Unfortunately the presentation is not always easy to follow, so this article presents an alternative approach.

Suppose that a pool with runners has amounts

X1

bet by oneself, and amounts

S1

bet by the rest of the market. The true probabilities of the runners winning the race are

P1

calculated from the previous form of the runners.

Let   be the total proportion of the pool that is paid out. Then the optimal bet proportions are given by

Equation 1

where

Equation 2

where the summation notation on the top means ‘add all values of Sifor which the expected return from runneriis negative’, and the corresponding sum on the bottom means ‘add all values ofPifor which the expected return from runneriis positive’.

 

Worked example

Consider a race with 8 dogs. The win pool sizes and probabilities are as shown in the following table:

1  $  148.00 0.1204
2  $  151.00 0.1741
3  $    36.00 0.0668
4  $  105.00 0.2048
5  $  135.00 0.2474
6  $    39.00 0.0500
7  $    22.00 0.0393
8  $    75.00 0.0974

The total pool size is $711, the track take-out is 19%, and the rebate is 9%. We assume that the pool is will grow by 15% from the time of placing any bets to when final dividends are calculated, and we decide to use 40% of the recommended Isaacs stake to reduce volatility. What bet sizes give the highest return?

The first step is to work out which dogs have a positive expectation, once take-out and rebate are taken into account. These are dogs for whichThe total pool size is $711, the track take-out is 19%, and the rebate is 9%. We assume that the pool is will grow by 15% from the time of placing any bets to when final dividends are calculated, and we decide to use 40% of the recommended Isaacs stake to reduce volatility. What bet sizes give the highest return?

Equation 3

where, for dog i,

Piis the probability of the dog winning;

Siis the amount staked in the public pool on this dog winning;

ESiis the total amount staked in the pool (i.e., the pool size);

T is the track take-out;

and is the track rebate. The results are:

Dog LHS RHS Bet
1 0.578239 1.132631 0
2 0.819606 1.132631 0
3 1.319096 1.132631 1
4 1.386482 1.132631 1
5 1.302859 1.132631 1
6 0.910893 1.132631 0
7 1.27093 1.132631 1
8 0.922943 1.132631 0

In this instance, dogs 3, 4, 5 and 7 have positive expectations.

The next step is to form the expression

Equation 4

 

This is straightforward using the results of the earlier calculation. Note that the factor   refers to the track take-out and does not include any rebate.

Dog sumSi sumPi
1 148.00 0
2 151.00 0
3 0.00 0.06679
4 0.00 0.204755
5 0.00 0.247378
6 39.00 0
7 0.00 0.039326
8 75.00 0
TOTAL 413.00 0.56

In this example, Y has a value of sqrt (0.9*$413/ 1-(0.9*0.56)) = 23.77.

The last step is to calculate the size of the actual bets using

Equation 5

The value of this expression will only be positive for the dogs with positive expectation. Again using the values above, the amounts to be staked on dogs 3, 4, 5, 7 are ($6.38, $21.73, $22.95, $3.42).

The last step is to take the increase in the pool size and the reduction factor into account. We expect the pool size to increase in size by 15% before the race starts, so all bets are multiplied by 1.15. Also, we have decided to bet 40% of the recommended Isaacs stake, so all bets are also multiplied by 0.40. The resulting bet sizes are ($2.94, $10.00, $10.56, $1.57).

 

[1] Isaacs, R. (1953)“Optimal Horse Race Bets” American Mathematical Monthly Vol. 60, No.5, (May, 1953), pp. 310-315, (May, 1953), pp. 310-315

About Andrew Colin

Andrew holds a PhD in Mathematics from the University of St Andrews, is a Fellow of the Institute of Mathematics and its Applications, and holds Chartered Mathematician accreditation. Andrew is also Adjunct Professor at the Faculty of Business at the University of Tasmania.
Andrew’s previous experience includes design and implementation of forecasting and arbitrage algorithms for soccer, horse racing and greyhound racing. He has also working in defence, academia, as a proprietary trader and a bond fund manager.
No Thoughts on Optimal bets for dog and horse races

Leave A Comment