Assume that a gambler has the possibility to bet a fraction of his capital in the outcome of a specific event. The Kelly criterion first presented in [1] and summarized below find the that maximizes the exponential rate of growth of the gambler’s capital under different scenarios, which is equivalent to maximizing period by period the expected log utility based on the current capital.

Discussion on why this choice of optimization makes sense was formally discussed in [2] and might be the subject of a future post. Intuitively, it makes sense to use this criterion if you bet regularly and reinvest your profits.

** Exponential rate of growth **

Lets define a quantity called the exponential rate of growth of the gambler’s capital, where

and is the gambler’s capital after bets, is his starting capital, and the logarithm is to the base two. is the quantity we want to maximize.

** Perfect knowledge **

In the case of perfect knowledge, the gambler would know the outcome of the event before anyone else and would be able to bet his entire capital at each bet. Then, and .

** Binary events **

Consider now a binary event where the gambler has a probability of success and a probability of failure. In this case the gambler would go broke for large with probability if he betted all his capital in each bet, even though the expected value of his capital after bets is given by

Because of that, let us assume that the gambler will bet a fraction of his capital each time. Then

where and are the number of wins and losses after bets. Following the definition given in Eq. (1), it can be shown that

Maximizing Eq. (2) with respect to gives

where is called the edge.

If the payoff is for a win and for a loss, then the edge is , the odds are , and

** Multiple outcome events **

Lets now consider the case where the event has more than two possible outcomes, not necessarily equally likely.

** – Fair odds and no “track take” **

Lets first consider the case of fair odds and no “track take”, that is

where is the probability of observing the outcome in a given event, as estimated by the entity offering the odds.

Consider to be the fraction of the gambler’s capital that he decides to bet on based on his belief of the probability of observing the outcome in a given event. The gambler’s estimated probability for an outcome will be denoted by .

Since there is no “track take”, there is no loss in generality in assuming that

That is, the gambler bets his total capital divided among the possible outcomes.

In this case, [1] have shown that

That is, the gambler should allocate his capital according to how likely he thinks each outcome is.

** – Unfair odds and no “track take” **

In this case

but are not necessarily equal to . Since there is no track take we can still consider .

Here, the value of that maximizes is again given by . Interesting conclusions can be taken from this result:

- As with the case of fair odds, is maximized by putting . That is, the gambler ignores the posted odds in placing his bets!
- Subject to , the value of is minimized when . That is, any deviation from fair odds helps the gambler.

** – When there is a “track take” **

In case there is a track take, it can no longer be assumed that . Let be the fraction not bet by the gambler.

The maximization process derived in [1] may be summarized as follows:

- (a) Permute indices so that
- (b) Set b equal to the minimum positive value of
- (c) Set . The will sum to .

It should be noted that if for all no bets are placed. But if the largest some bets might be made for which , i.e. the expected gain is negative. This violates the criterion of the classical gambler who never bets on such events.

**References:**

[1] Kelly, J. L. (1956). A new interpretation of information rate. Information Theory, IRE Transactions on, 2(3), 185-189.

[2] Breiman, L. (1961). Optimal gambling systems for favorable games.

[3] MacLean, L. C., Thorp, E. O., Ziemba, W. T. (Eds.). (2011). The Kelly capital growth investment criterion: Theory and practice (Vol. 3). world scientific.

I have had Mathematical Programming and Games by Edward Kaplan from the library

and I am hoping I while apply wind myself up and try the Simplex method with multiple objective functions and sensitivity analysis which I am hoping is another way at portfolio selection.

Thanks for Kelly article is new and very interesting for me.

regards,

Jeremy

I don’t know why I said portfolio selection, it is not clear, In betting on horse races there is a choice of races, horses and bet types, so what I meant is optimising the choice of those three factors given the prior (subjective) selections made for a particular race day.

I would like to say more but I fear I will expose myself to contempt. I have not found a way out of the subjectivity problem and my suspicion is that there is not one. I am not sure what this means.

Jeremy