Welcome to Scribd, the world's digital library. Read, publish, and share books and documents. See more
Download
Standard view
Full view
of .
Save to My Library
Look up keyword
Like this
5Activity
0 of .
Results for:
No results containing your search query
P. 1
St. Petersburg Portfolio Games

St. Petersburg Portfolio Games

Ratings: (0)|Views: 503|Likes:
Published by TraderCat Solaris

More info:

Published by: TraderCat Solaris on Jun 26, 2010
Copyright:Attribution Non-commercial

Availability:

Read on Scribd mobile: iPhone, iPad and Android.
download as PDF, TXT or read online from Scribd
See more
See less

10/14/2012

pdf

text

original

 
St. Petersburg Portfolio Games
aszl´o Gy¨orfi
and eter Kevei

Department of Computer Science and Information TheoryBudapest University of Technology and EconomicsMagyar Tud´osok k¨or´utja 2., Budapest, Hungary, H-1117
gyorfi@szit.bme.hu
Analysis and Stochastics Research GroupHungarian Academy of SciencesAradi ertan´uk tere 1, Szeged, Hungary, H-6720
kevei@math.u-szeged.hu
Abstract.
We investigate the performance of the constantly rebalancedportfolios, when the random vectors of the market process
{
X
i
}
are inde-pendent, and each of them distributed as (
X
(1)
,X
(2)
,...,X
(
d
)
,
1),
d
1,where
X
(1)
,X
(2)
,...,X
(
d
)
are nonnegative iid random variables. Undergeneral conditions we show that the optimal strategy is the uniform:(1
/d,...,
1
/d,
0), at least for
d
large enough. In case of St.Petersburg com-ponents we compute the average growth rate and the optimal strategyfor
d
= 1
,
2. In order to make the problem non-trivial, a commission fac-tor is introduced and tuned to result in zero growth rate on any individ-ual St. Petersburg components. One of the interesting observations madeis that a combination of two components of zero growth can result in astrictly positive growth. For
d
3 we prove that the uniform strategy isthe best, and we obtain tight asymptotic results for the growth rate.
1 Constantly Rebalanced Portfolio
Consider a hypothetical investor who can access
d
financial instruments (asset,bond, cash, return of a game, etc.), and who can rebalance his wealth in eachround according to a portfolio vector
b
= (
b
(1)
,...,b
(
d
)
). The
j
-th component
b
(
j
)
of 
b
denotes the proportion of the investor’s capital invested in financialinstrument
j
. We assume that the portfolio vector
b
has nonnegative componentsand sum up to 1. The nonnegativity assumption means that short selling is notallowed, while the latter condition means that our investor does not consumenor deposit new cash into his portfolio, but reinvests it in each round. The setof portfolio vectors is denoted by
Δ
d
=
b
= (
b
(1)
,...,b
(
d
)
);
b
(
j
)
0
,
d
j
=1
b
(
j
)
= 1
.
The first author acknowledges the support of the Computer and AutomationResearch Institute of the Hungarian Academy of Sciences.

The work was supported in part by the Hungarian Scientific Research Fund, GrantT-048360.
R. Gavald`a et al. (Eds.): ALT 2009, LNAI 5809, pp. 83–96,2009.c
Springer-Verlag Berlin Heidelberg 2009
 
84 L. Gy¨orfi and P. Kevei
The behavior of the market is given by the sequence of return vectors
{
x
n
}
,
x
n
= (
x
(1)
n
,...,x
(
d
)
n
)
,
such that the
j
-th component
x
(
j
)
n
of the return vector
x
n
denotes the amount obtained after investing a unit capital in the
j
-th financialinstrument on the
n
-th round.Let
0
denote the investor’s initial capital. Then at the beginning of the firstround
0
b
(
j
)1
is invested into financial instrument
j
, and it results in return
0
b
(
j
)1
x
(
j
)1
, therefore at the end of the first round the investor’s wealth becomes
1
=
0
d
j
=1
b
(
j
)1
x
(
j
)1
=
0
b
1
,
x
1
,
where
·
,
·
denotes inner product. For the second round
b
2
is the new portfolioand
1
is the new initial capital, so
2
=
1
·
b
2
,
x
2
=
0
·
b
1
,
x
1
·
b
2
,
x
2
.
By induction, for the round
n
the initial capital is
n
1
, therefore
n
=
n
1
b
n
,
x
n
=
0
n
i
=1
b
i
,
x
i
.
(1)Of course the problem is to find the optimal investment strategy for a longrun period, that is to maximize
n
in some sense. The best strategy dependson the optimality criteria. A naive attitude is to maximize the expected returnin each round. This leads to the risky strategy to invest all the money intothe financial instrument
j
, with
E
(
j
)
n
= max
{
E
(
i
)
n
:
i
= 1
,
2
,...,n
}
, where
X
n
= (
(1)
n
,
(2)
n
,...,X 
(
d
)
n
) is the market vector in the
n
-th round. Since therandom variable
(
j
)
n
can be 0 with positive probability, repeated application of this strategy lead to quick bankrupt. The underlying phenomena is the simplefact that
E
(
n
) may increase exponentially, while
n
0 almost surely. A moredelicate optimality criterion was introduced by Breiman[3]: in each round wemaximize the expectation
E
ln
b
,
X
n
for
b
Δ
d
. This is the so-called
log-optimal portfolio
, which is optimal under general conditions[3].If the market process
{
X
i
}
is memoryless, i.e., it is a sequence of independentand identically distributed (i.i.d.) random return vectors then the log-optimalportfolio vector is the same in each round:
b
:= argmax
b
Δ
d
E
{
ln
b
,
X
1
}
.
In case of constantly rebalanced portfolio (CRP) we fix a portfolio vector
b
Δ
d
.In this special case, according to(1) we get
n
=
0
ni
=1
b
,
x
i
, and so theaverage growth rate of this portfolio selection is1
n
ln
n
=1
n
ln
0
+1
n
n
i
=1
ln
b
,
x
i
,
 
St. Petersburg Portfolio Games 85
therefore without loss of generality we can assume in the sequel that the initialcapital
0
= 1.The optimality of 
b
means that if 
n
=
n
(
b
) denotes the capital afterround
n
achieved by a log-optimum portfolio strategy
b
, then for any portfoliostrategy
b
with finite
E
{
ln
b
,
X
1
}
and with capital
n
=
n
(
b
) and for anymemoryless market process
{
X
n
}
1
,lim
n
→∞
1
n
ln
n
lim
n
→∞
1
n
ln
n
almost surelyand maximal asymptotic average growth rate islim
n
→∞
1
n
ln
n
=
:=
E
{
ln
b
,
X
1
}
almost surely.The proof of the optimality is a simple consequence of the strong law of largenumbers. Introduce the notation
(
b
) =
E
{
ln
b
,
X
1
}
.
Then the strong law of large numbers implies that1
n
ln
n
=1
n
n
i
=1
ln
b
,
X
i
=1
n
n
i
=1
E
{
ln
b
,
X
i
}
+1
n
n
i
=1
(ln
b
,
X
i
E
{
ln
b
,
X
i
}
)=
(
b
) +1
n
n
i
=1
(ln
b
,
X
i
E
{
ln
b
,
X
i
}
)
(
b
) almost surely.Similarly,lim
n
→∞
1
n
ln
n
=
(
b
) = max
b
(
b
) almost surely.In connection with CRP in a more general setup we refer to Kelly [8]andBreiman[3].In the following we assume that the i.i.d. random vectors
{
X
i
}
, have thegeneral form
X
= (
(1)
,
(2)
,...,X 
(
d
)
,
(
d
+1)
), where
(1)
,
(2)
,...,X 
(
d
)
arenonnegative i.i.d. random variables and
(
d
+1)
is the cash, that is
(
d
+1)
1,and
d
1. Then the concavity of the logarithm, and the symmetry of the first
d
components immediately imply that the log-optimal portfolio has the form
b
= (
b,b,...,b,
1
db
), where of course 0
b
1
/d
. When does
b
= 1
/d
correspond to the optimal strategy; that is when should we play with all ourmoney? In our special case
has the form
(
b
) =
E
ln
b
d
i
=1
(
i
)
+ 1
bd

.

Activity (5)

You've already reviewed this. Edit your review.
1 hundred reads
lince82 liked this
indisateesh liked this
James liked this

You're Reading a Free Preview

Download
/*********** DO NOT ALTER ANYTHING BELOW THIS LINE ! ************/ var s_code=s.t();if(s_code)document.write(s_code)//-->