MPI-INF Logo
Campus Event Calendar

Event Entry

What and Who

Online Algorithms to Determine Large Portfolios

Christoph Hundack
Max-Planck-Institut für Informatik, Max-Planck-Institut für Informatik, Saarbrücken
Seminar des Graduiertenkollegs
AG 1, AG 2  
AG Audience
-- Not specified --

Date, Time and Location

Monday, 23 November 98
16:00
-- Not specified --
45
015
Saarbrücken

Abstract

Markowitz' mean--variance analysis is the most frequently used

technique for portfolio selection monitoring reward and risk.
In a seminal paper Cover (1991) proposed a different understanding
of portfolio risk by measuring it wrt a reference portfolio.
His online trading strategy results in a portfolio P which has
asymptotically the same exponential growth rate as the reference
portfolio. The offline reference portfolio is chosen as the optimal
portfolio with a fixed distribution determined ex poste,
ie with full knowledge of the stock rates within the considered
time period. The log-optimal portfolio P is called universal
as no statistical assumptions on market parameters are made.

Requiring extensive computation this strategy has been limited
to universal portfolios containing up to three assets. By using new
approaches this strategy in extended to universal portfolios of
arbitrary size.

Results on the German stock market (1987-1997, daily basis) confirm
that the universal portfolio outperforms the value-line-index
and show its compatibility with the DAX-30.

Contact

Ülkü Coruh
0681/9325-526
--email hidden
passcode not visible
logged in users only