06.06.2013 Views

STOCHASTIC

STOCHASTIC

STOCHASTIC

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

(b) Interpret this assumption. Is it strong or weak?<br />

(c) Show that Breiman's assumption that the random returns are finite and bounded<br />

away from zero is sufficient but not necessary for the satisfaction of the boundedness<br />

assumption.<br />

(d) Use the Chebychev inequality to prove that for independent returns, a max E log VN<br />

criterion has a modified version of Property 1 (where convergence in probability replaces<br />

almost sure convergence) stated in Thorp's paper. Assume that all feasible strategies<br />

are bounded.<br />

(e) Use the Chebychev inequality to prove that for dependent returns, a max E log VN<br />

criterion has the modified Property 1. Assume that all strategies under consideration<br />

are significantly different and bounded, and that the return distributions are independent<br />

of the investor's strategy.<br />

(f) What can be said about Property 2 in Thorp's paper ?<br />

•24. Refer to the papers by Thorp and Breiman.<br />

(a) Attempt to prove that Thorp's Property 2 does hold for Breiman's model.<br />

(b) Attempt to prove that for a sequence of investment choices Thorp's Property 1<br />

holds if and only if Property 2 holds.<br />

(c) Assume that the investor's initial wealth is the positive integer M, that only integervalued<br />

returns are possible, and that the X, must be integers. Suppose In* is a maximum<br />

expected log strategy. Attempt to show that Theorems 1-3 apply to 1N*.<br />

(d) Does Property 2 hold for this model ?<br />

25. (The Kelly criterion and expected utility) Suppose an investor's utility function is<br />

WT" over period T's wealth WT, where a < 1 and a ^ 0. Let the gross rate of return in each<br />

period t be r,, where r, = £ rt Xit, the X„ are the relative investment allocations in period t,<br />

and the rt are the gross rates of return for the individual investments assumed to be independently<br />

and identically distributed in time. Suppose Wa > 0 is the investor's initial wealth;<br />

then his wealth at time 7" is<br />

WT= Waf[rx.<br />

t-i<br />

(a) Show that one may maximize WT for fixed r, by choosing the X„ so that in each<br />

period they maximize the expected logarithm of one period return, namely,<br />

max log ][>,*„, s.t. ~£Xlt = 1, X„ 5; 0.<br />

i i<br />

(b) Show that the optimal strategy, say A", is stationary in time, i.e., X' — (Xlt, ...,Xm)<br />

for all t.<br />

(c) Show that with probability 1 the return under the X' strategy is at least as high as<br />

under the strategy X p , where X p is the solution to<br />

max£Q>,AY)

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!