Seventh OCCF City Seminar

sponsored by SUN, IBM, Reuters, NAG, and Moneyline Telerate

Sun Microsystems Ltd, Regis House,
45 King William Street, London EC4R 9AN
(Near Monument tube station)

Tuesday February 28th 2006, 4-7pm

Slides available -- see below


Agenda:

1600 Coffee
1615 Fast adjoint calculation of Monte Carlo Greeks.
Prof Mike Giles, Oxford University
1700 Generalised mean variance analysis and robust portfolio construction.
Dr Stephen Wright, Equities Research Department, UBS Limited
1745 Break
1800 Continuity corrections for discretely sampled options.
Dr Sam Howison, Oxford University
1845 Close. Drinks and Canapés.

Abstracts:

Fast adjoint calculation of Monte Carlo Greeks -- Prof Mike Giles, Oxford University

In this talk, Mike will explain how the use of adjoint methods makes it possible to compute many Greeks for a total computational cost comparable to the payoff evaluation. Numerical results will be shown for deltas and vegas from a LIBOR market model example.

Slides now available as a PDF file.

Generalised mean variance analysis and robust portfolio Construction -- Dr Stephen Wright, Equities Research Department, UBS Limited

In the talk Stephen will discuss the problems that occur with the use of "textbook" mean variance approaches to portfolio construction. He will outline some of the alternative approaches that have been developed to address these issues, and will then go into detail on the rank approach to diversification that he has recently developed with Steve Satchell.

This talk is based on a paper of the same title, co-authored with Steve Satchell, which appeared in Advances in Portfolio Construction and Implementation, published in 2003 by Butterworth and Heinemann.

Slides now available as a PPT file.

Continuity corrections for discretely sampled options -- Dr Sam Howison, Oxford University

Sam will be looking at the BGK `continuity correction' for discretely sampled options, using the method of matched asymptotic expansions. He will show that the correction is very generally applicable, and how to use it for Bermudan options.

Slides now available as a PDF file.


Sixth OCCF City Seminar

sponsored by IBM, Reuters, SUN, NAG, and Moneyline Telerate

Wednesday May 25th 2005, 4-7pm

Slides available -- see below


Agenda:

1600 Coffee
1615 Inter-pattern speculation: beyond minority, majority and $-games.
Damien Challet, Oxford University
1700 Modelling the performance of hedge fund managers.
Anthony Ledford, MAN Investments
1745 Break
1800 Computing Greeks with Monte Carlo methods.
Mike Giles, Oxford University
1845 Close. Drinks and Canapés.

Abstracts:

Inter-pattern speculation: beyond minority, majority and $-games -- Damien Challet, Oxford University

A new model of financial markets is proposed, based on the sequential and inter-temporal nature of trader-trader interaction. In this pattern-based speculation model, the traders open and close their positions explicitely. Information ecology can be precisely characterised, and is strikingly similar to that of the Minority Game. Naive and sophisticated agents are shown to give rise to very different phenomenology.

Slides now available as a PDF file.

Modelling the performance of hedge fund managers -- Anthony Ledford, MAN Investments

Collaborators: Darren Upton and Mike Robinson, Man Investments

To model the performance of hedge fund managers it is desirable to construct a model that captures many features that are salient in financial datasets. These features include heavy tailed asymmetric distributions, dynamic variances, a mean that is a function of variance, and dependence between different managers. Each of these topics is briefly discussed with reference to a dataset that contains the performance of 4000 managers over the last 15 years.

We propose a multivariate time series model that accommodates these key features and allows for missing values. We fit it using maximum likelihood, examine model adequacy, and then briefly discuss some practical implications.

Slides now available as a PDF file.

Computing Greeks using Monte Carlo methods -- Mike Giles, Oxford University

In this talk I will describe methods of performing Monte Carlo pathwise sensitivity calculations to determine Greeks such as Delta and Gamma. This often requires regularisation (i.e. smoothing) of the payoff function, but a recursive correction procedure can be used to reduce the bias that this introduces. One is then left with a tradeoff between high smoothing which introduces more bias, and low smoothing which leads to a higher Monte Carlo sampling error. Numerical results will show that this sampling error can be reduced by using adaptive stratified sampling.

Slides now available as a PDF file.


Fifth OCCF City Seminar

sponsored by SUN, Reuters, IBM, NAG, and Moneyline Telerate

Wednesday September 22nd 2004, 4-7pm

Slides available -- see below


Agenda:

1600 Coffee
1615 Computational strategies for high-dimensional option pricing problems.
Christoph Reisinger, Oxford University
1700 The Foreign Exchange Market and the Growing Role of Quantitative Analysis.
Stacy Williams, HSBC
1745 Break
1800 Correlations between Foreign-Exchange Returns - A Minimal Spanning Tree Approach.
Omer Suleman, Oxford University
1845 Close. Drinks and Canapés.

Abstracts:

Computational strategies for high-dimensional option pricing problems -- Christoph Reisinger, Oxford University

The pricing of options depending on a large number of risk factors is often believed to be computationally unfeasible in the PDE context. Recent numerical developments suggest that this is not necessarily true. Therefore the aim of this talk is to give a survey of techniques, which are currently being developed in order to solve efficiently those high-dimensional PDEs. The focus will not only be on European options, but also examples with American or Bermudan exercise will be considered.

Vital for this approach is the interplay of a number of sophisticated analytical and numerical techniques including asymptotic expansions for the construction of the solution from lower-dimensional auxiliary problems, sparse grid and extrapolation techniques for an optimal choice of approximation spaces, multigrid methods for the fast solution of the discrete equations as well as the efficient parallel implementation of the algorithms on distributed memory machines.

Slides available as a PDF file.

The Foreign Exchange Market and the Growing Role of Quantitative Analysis -- Stacy Williams, HSBC

The foreign exchange market is the largest market in the world with a turnover of around $1.2 trillion per day. It is a highly competitive and transparent market in which it is becoming increasingly difficult for market making banks to differentiate themselves. By providing innovative research to their clients, banks are striving to gain market share, and it is in the area of quantitative analysis, in particular, where the greatest advances are being made. Here, we examine some of the latest available research, and look at the ever critical role that technology plays in enabling such quantitative analysis to be performed.

Slides available as a PDF file.

Correlations between Foreign-Exchange Returns: A Minimal Spanning Tree Approach -- Omer Suleman, Oxford University

Minimal Spanning Trees have recently been introduced as a tool for studying the dynamics of financial markets. Tree based on daily returns have been used in the equity market to determine their long term behaviour. In our work at OCCF Minimal Spanning Trees are constructed from the cross-rate returns in the currency market using high-frequency data. The stability and time dependence of the trees is investigated and attempts are made at determining which currencies are 'in play' given the positions of various currency nodes in the tree.

Research performed in association with M. McDonald, S. Williams (HSBC), N.F. Johnson and S. Howison.

Slides available as a Powerpoint file.


Fourth OCCF City Seminar

sponsored by SUN, Reuters, IBM, NAG, and Moneyline Telerate

Thursday 11th March 2004, 4-7pm

Slides available -- see below


Agenda:

1600 Coffee
1615 Robust Markovitz Portfolio Optimisation.
Raphael Hauser, Oxford University
1645 Advanced Finite Difference Schemes.
William Shaw, Oxford University
1745 Break
1800 Parallel Computing and Applications in Finance.
Cyril Godart, BNP Paribas
1900 Close. Drinks and Canapés.

Abstracts:

Robust Markovitz Portfolio Optimisation -- Raphael Hauser, Oxford University

Markovitz portfolio optimisation assumes that the covariance matrix for the random returns of a number of different assets is known. In practice this matrix has to be estimated from historical data, and in the typical case where a few of its eigenvalues are close to zero, the results obtained under the Markovitz model are sensitive to errors in the covariance estimates. I will show a very simple procedure to robustify the results based on treating the covariance estimates as confidence intervals and considering all the resulting risk structures jointly. This leads to a semidefinite programming problem which can be solved efficiently via software available in the public domain.

Slides available as a PDF file.

Advanced Finite Difference Schemes -- William Shaw, Oxford University

This talk will present some of the exotic two and three time level systems. An analysis of their theoretical properties will be combined with a presentation of their practical benefits for financial systems.

Slides available as a PDF file.

Parallel Computing and Applications in Finance -- Cyril Godart, BNP Paribas

The presentation intends to give a survey of the potential applications of Parallel Computing in Finance. From calibration of interest rate models to speed-up of huge Monte-Carlo simulation, they are to be found everywhere. Nevertheless, there are some theoritical and practical difficulties when it comes to implementing such a platform. We will review them, present a theoretical framework to tackle the task and give concrete examples of how efficient implementations can be. The presentation is intended for a general audience and requires no prior knowledge in Finance or Parallel Computing.

Slides available as a PDF file.


Third OCCF City Seminar

sponsored by Reuters, IBM, SUN, NAG, and Moneyline Telerate

Wednesday 18th June 2003, 4-7pm

Slides available -- see below


Agenda:

1600 Coffee
1615 OCCF - The Realtime Grid.
Rupert Brown, OCCF/Reuters
1645 A parametric model for liquidity effects in derivative pricing and hedging.
Sam Howison, OCCF/Oxford
1745 Break
1800 Invited Presentation: Using Bayesian Mixed Estimation Methods to implement Enhanced Index strategies.
Alan Scowcroft, Managing Director, Global Head of Equities Quantitative Research, UBS Warburg
1900 Close. Drinks and Canapés.

Abstracts:

OCCF: The Realtime Grid -- Rupert Brown, OCCF/Reuters

A guided tour of the major technical components of the OCCF grid installation as well as insight into the resources used to build and maintain the installation day to day. There will also be a sneak peek at the likely future technology roadmap as the centre evolves.

Slides available as a Powerpoint file.

A parametric model for liquidity effects in derivative pricing and hedging -- Sam Howison, OCCF/University of Oxford

The work described in this talk, jointly with David Bakstein, deals with a model for pricing and hedging derivatives in a market where trading incurs transaction costs and causes price slippage. We formulate discrete and continuous-time models, analyse them approximately and numerically, and show that the model can be calibrated to (equity) market data.

Using Bayesian Mixed Estimation Methods to implement Enhanced Index strategies -- Alan Scowcroft, Managing Director, Global Head of Equities Quantitative Research, UBS Warburg

The application of Bayesian Mixed Estimation techniques is relatively well understood for problems involving a small number of assets. However, its extension to benchmarks of thousands of assets presents some interesting computational challenges. The approach will be illustrated in the context of combining portfolio forecasts to efficiently implement active quantitative strategies. Consideration will also be given to alternative approaches to robust optimisation. The talk will conclude with a discussion of some recent advances in the application of Mixed Estimation.

Slides available as a PDF file.


Second OCCF City Seminar

sponsored by IBM, SUN, NAG, Reuters, and Moneyline Telerate

Wednesday, February 12, 2003, 4-7pm

Slides available -- see below


Agenda:

1600 Coffee
1615 Introductions
1645 What makes a financial market 'tick': The complexities of financial market dynamics.
Prof. Neil Johnson, Director OCCF
1745 Break
1800 eScience and grid computing.
Prof. Mike Giles, Director OCCF
1900 Close. Drinks and Canapés.

Abstracts:

What makes a financial market 'tick': The complexities of financial market dynamics -- Prof. Neil Johnson

It has been said that being a successful investor is like being a successful burglar. Most burglars know how to get into a building, and what to take - but only the successful ones know how, and most importantly when, to get out. `Time' is therefore crucial in financial markets. Since our goal at OCCF is to understand real-world markets as opposed to idealized ones, time also takes centre-stage in our research. Time relates to dynamics, and it is this evolving complex-system dynamics which underpins our scientific interest in financial markets. The following questions provide the focus of this talk: How do financial markets really behave? Why do they behave in the way that they do? What can we do to minimize risk, given this behaviour? Most, if not all, standard finance treatments make some kind of apparently innocuous assumptions about market dynamics. For example they assume that the markets are in some kind of steady state or represent a stationary process, and that there are no implicit temporal correlations - or at best, that these correlations are of a specific type. As we show in this talk, these assumptions can give misleading answers to practical problems such as minimizing financial risk, coping with extreme events such as crashes or drawdowns, and pricing derivatives in non-ideal markets. Having said this, standard finance theory usually works. But it is the `usually' that we are interested in or rather the `unusually'. This talk focuses on how and why real financial markets deviate from the standard finance theory paradigm of random-walk behaviour, and the consequences of such deviation. In particular, we are interested in the tails of the distribution of price returns, and in the dynamics induced by crowd-like behaviour in markets. The consequences for managing risk will also be discussed.

Slides available as a PDF file.

eScience and Grid Computing -- Prof. Mike Giles

The UK has invested 120M in a 3-year academic initiative called eScience, and is about to invest even more in a two-year follow-on. In this talk, I will give an overview of this initative to develop tools and infrastructure to promote large-scale scientific collaborations, and relate it to work on Grid Computing in the US and elsewhere. I will also discuss the possibility for projects in the financial sector in the next round of funding.

Slides available as a PDF file.


Inaugural OCCF City Seminar

Wednesday 25 September 2002, 4-7pm

Slides available -- see below


Agenda:

1600 Coffee
1615 Introduction to the OCCF
Prof. Mike Giles, Director OCCF
1630 The place of academic research in real-world finance.
Nigel Webb, Partner (Capital Markets, Financial Services)
Deloitte and Touche Consulting
1700 Numerically Intensive Computing in Finance.
Prof. Mike Giles, Director OCCF
1750 Break
1810 Monte Carlo for American Options.
Dr. Jeff Dewynne, Director OCCF
1900 Close. Drinks and Canapés.

Abstracts:

Numerically Intensive Computing in Finance -- Mike Giles

This presentation will give an overview of current trends in numerically intensive computing for financial applications such as options pricing. Starting with the underlying hardware, from clusters of commodity machines up to very large shared-memory machines, I will then discuss the loosely-coupled or tightly-coupled nature of different parallel applications, and how they are implemented using distributed resource management and standards such as OpenMP and MPI. The material will be illustrated through an application concerning the valuation of a gas storage facility, modelled with both finite difference and Monte-Carlo methods.

Slides available as a PDF file.

Monte Carlo Methods for American Options -- Jeff Dewynne

There are highly effective grid-based numerical methods for American options in low dimensions. The numerical valuation of an American option in low dimensions is typically as effective as its European equivalent. Monte Carlo methods are the only effective numerical means of pricing European options in high dimensions. Unfortunately they can not be used, easily, for American options which means that, effectively, there is no means of numerically valuing an American option in high dimensions. In this talk I will review some of the attempts to value American options using Monte Carlo methods (Broadie, Glasserman and Rogers) and describe in somewhat more detail the approach we are pursuing in OCIAM and the OCCF. The key idea is that you can use Monte Carlo if you know what the optimal exercise strategy is. Our approach is to use techniques from fluid dynamics (specifically, matched asymptotic expansions) in order to obtain asymptotic expansions for the optimal exercise strategy and then perform simulations using this asymptotic approximation.

Slides available as a PDF file.