Copyright © 2009 by Daryl Roxburgh. All rights reserved.
Doi:10.1016/B978-0-12-374952-9.00003-8.

2010

Optimal solutions for optimization

in practice

Daryl Roxburgh , Katja Scherer and Tim Matthews

3

Executive Summary

Optimization has been one of the principal techniques to allow quantitative

investors to build portfolios. It is not the only technique and there are styles of

investment that are not automatically suitable for optimization. Historically, opti-

mization was subject to a detailed criticism by disgruntled users due to e.g. the

solutions being excessively sensitive to unstable inputs. It should be noted that

optimization has experienced a renaissance in the last five or so years as many

high-frequency hedge funds have used it to build portfolios on a daily, or even

intra-daily basis. Furthermore, techniques such as Robust and the innovative

gain-loss optimization (GLO) have been developed that have addressed prob-

lems associated with alpha forecasting errors and the non-normality of returns;

we shall discuss these in this chapter together with methods to combine mean-

variance and gain-loss to benefit from both their strengths. We conclude with a

practical application of such techniques and with a particular focus on the area

of Charities and Endowments where we consider investments with infinite invest-

ment horizons.

3.1 Introduction

Optimization has been one of the principal techniques to allow quantitative
investors to build portfolios. It is not the only technique and there are styles of
investment that are not automatically suitable for optimization. Historically,
optimization was subject to a detailed criticism by disgruntled users due to the
solutions being excessively sensitive to errors in forecast of alpha; we will not
discuss this in any detail as it is covered elsewhere in the book. It should be
noted that optimization has experienced a renaissance in the last 5 or so years
as many high-frequency hedge funds have used it to build portfolios on a daily,
or even intradaily basis. Furthermore, techniques have been developed that have
addressed the problems associated with alpha measurement; we shall discuss
these later in the chapter.