Scribd
Upload
31 views2 pages

Optimal Gambling Systems

The document discusses optimal gambling systems and strategies. It presents three theorems about the fortunes of gamblers using different betting strategies under varying conditions of wager odds and probabilities of winning or losing bets. The theorems address whether a gambler's fortune will converge or diverge, the optimal strategy for maximizing the chance of reaching a goal amount, and the conditions under which the maximum probability of exceeding a starting amount equals that amount.

Uploaded by

cmcclos
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
31 views2 pages

Optimal Gambling Systems

The document discusses optimal gambling systems and strategies. It presents three theorems about the fortunes of gamblers using different betting strategies under varying conditions of wager odds and probabilities of winning or losing bets. The theorems address whether a gambler's fortune will converge or diverge, the optimal strategy for maximizing the chance of reaching a goal amount, and the conditions under which the maximum probability of exceeding a starting amount equals that amount.

Uploaded by

cmcclos
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 2

OPTIMAL GAMBLING SYSTEMS

BY LESTER E. DUBINSANDLEONARDJ. SAVAGE


IJNIVERSITY OF CALIFORNIA AT BERKELEY AND UNIVERSITY OF MICHIGAN

Communicated by J. L. Doob, October 19, 1960

Until recently, the mathematics of optimal gambling has been primarily con-
cerned with the formulation and proof of variants of the following fact. The
expected terminal fortune of a gambler, whose fortune is always nonnegative,
engaged in any succession of gambles that are fair or unfair to him cannot be larger
than his initial fortune.
These results, though important, do not say how well the gambler can do in
given unfair gambling situations, nor do they suggest, what gambling strategy is
best in his bad situation. These latter problems h'ave their own mathematical
interest, and the underlying ideas used in attacking. them can be applied to establish
various inequalities for wide classes of stochastic prdcesses of theoretical, and
perhaps practical, interest.
Some illustrations of our results in this area, to'be published elsewhere, are
here stated without proof in language that, for brevity, is somewhat informal.
Let r and w be two numbers less than 1. Suppose a gambler whose initial
fortune is positive can make a sequence of bets, on any one of which he can stake
any amount then in his possession. Each bet results in either a gain or a loss,
where the ratio of the possible gain to the possible loss is as i - r to r, the gain
occurs with probability w, and the loss occurs with probability 1 - w.
THEOREM 1. If w > r, there exists k, 0 < k < 1, such that a gambler who stakes
cy wheneverhis fortune is y will have his fortune convergealmost certainly to -+ o or 0,
according as 0 < c < k or c > k. If c = k, the lim sup and lim inf of Ihisfortunes
will almost certainly be + o and 0, respectively.
Let a certain positive G be a gambler's goal. Roughly speaking, a betting
strategy is optimal if it maximizes the probability that the gambler's fortune ever
attains the goal. A gambler plays boldly if he always stakes as much as possible
consistent with the condition that his fortune;never exceed G or become negative.
THEOREM 2. If w < r, bold play is optimal. Bold play is not the only optimal
strategy.
Let g be a random variable bounded below by -1 and of expectation 0-a fair
lottery, so to speak-and let gn, n = 1, 2, ..., be independent, each with the same
distribution as g. For any x, 0 < x < 1, consider any stochastic process f, n = 0,
1, 2, .. .the fortune of a gambler at time n--such that: (i) fo = x; (ii) f, > 0
for all n; (iii) fn+l = fn + Sn+lgn+l,where s--the stake at time n-is nonnegative
and may depend upon fo, ..., fn. Then o- = 'S, s,, ... is a strategy and p(o, x) is
the probability that for some n, fn > 1. Before stating a main result, we note
that well-known theorems easily imply: (iv) For all o-, p(a, x) < x; (v) if g is
bounded above and if its upper bound is attained with positive probability, then
there is a aosuch that p(a, x) = x; (vi) if g is bounded above but its upper bound
is not attained with positive probability, then for all aoand x, p(ac, x) < x but the
sup of p(u, x) over all aois x.
Suppose now that g is unbounded above. It is relatively easy to exhibit a g
1597

This content downloaded from 169.229.32.136 on Wed, 7 May 2014 14:49:14 PM


All use subject to JSTOR Terms and Conditions
1598 MATHEMATICS:L. P. EISENZART' Pioc. N. A. S.

such that U(x), the supremum of p(r,x) over all a, is x. It is much more difficult
to see, as Donald Ornstein was the first to do, that there exist g for which U(x) < x.
We now know more:
THEOREM 3. Suppose that g has mean 0 and that Ig has positive mean. Then
in order that U(x) = x it is necessary and sufficient that the following condition hold:
The limit inferior, as z approaches o, of (-z f g)( f g')-~ is 0.
9<z g<z
This research owes much to many, both for ideas and for financial support, as
will be explained in the full publication.

FIELDS OF UNIT VECTORS IN THE FOUR-SPACE OF GENERAL


RELATIVITY
BY LUTHER P. EISENHART

PRINCETON UNIVERSITY

October
Communicated 14, 1960
1. A previous paper' deals with the tangent vectors to the minimal geodesics of
the four-space of general relativity V4,which, according to Einstein, are the paths of
light.2 This means that the tensor gij in the fundamental equation
ds2 = gijdxidxj (1)
is not positive definite. The right-hand member of this equation stands for the
sum of terms as i and j take the values 1 to 4.
This convention is used throughout this paper, namely, that when within a term
the same letter enters as a superscript and subscript, the term stands for the sum
of terms as the index takes the values 1 to 4.
The present paper deals with unit vectors in the V4, that is,
Xl;"Xl;i = 1, a2;iX2;i = 1, X3;iX;i = 1, (2)

which are mutually orthogonal, that is,'

Xi;;i2; i = , X2; iX3;i = 0, X3; i;i = 0. (3)

2. We use the null vector Xi of the previous paper, that is,


iXi = 0, (4)
and require that the vectors Xl;i, X2;i,X,3;ibe orthogonal to the vector X,, that is,

Xl;i X = XiX,Xi = X2i;;i = ;iXi X:, = ;i


O0, i = 0. (5)
We put
Xl;i,j = XiXi;j, X2;i,j= XX2;j, X3;i,j= XiX3;j, (6)
where the terms on the left are the covariant derivatives of X; i, X2; i X3;i with respect
to xJ.4
This convention is used throughout this paper, namely, a component of a vector

This content downloaded from 169.229.32.136 on Wed, 7 May 2014 14:49:14 PM


All use subject to JSTOR Terms and Conditions

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy