C(omp)++ | A Portal for the Computational Finance Community: Excel

Since authors are, first and foremost, consumers of books, I wanted this edition to be worth buying again, as there’s nothing worse than a second edition that isn’t. However, large sections of the book are very much like the first, and so I’m very happy to be able to explain here what almost two years’ effort has been all about.

Holes have been filled and areas augmented, many of these arising from questions I have received since publication.

The example code class, cpp_xloper, that encapsulates Excel’s xloper data type, now wraps the call-backs Excel4 and Excel4v, allowing for faster, clearer and safer coding. There’s a skeleton XLL project containing a shell worksheet function and a shell command, providing a simple starting point.

There are new financial applications, including SABR and simple CMS derivatives. As before, these are there simply to illustrate how to encapsulate real problems with fast and flexible interfaces.

Microsoft Excel 2007 is the most important release, for XLL developers, since Excel 97. Much of the book’s new material explores the opportunities and challenges that this raises. Two key new features are multi-threaded workbook recalculation and a grid that is over 1,000 times larger than that of Excel 2003. Microsoft have written a new Excel C API and SDK which provides new data types that support the larger grids, and for the first time, long Unicode strings. Thread-safe XLL worksheet functions can be registered as such, enabling Excel 2007 to call them concurrently on multiple threads. This not only leads to faster recalculation on multi-processor and -core machines, but also XLLs that can harness highly-parallel servers. The application to Monte Carlo simulation is obvious.

Programming in a multi-threaded environment will be a new challenge for many XLL developers. Thread-safe programming techniques are carefully explained in an XLL context: where, when and how to use them.

I am very grateful to those who bought the first book and provided feedback. The second edition is a much better book in part because of them. I hope those who buy it will get as much out as I have tried to put in.

I’d be happy to respond to any questions or feedback you might have.

Steve

To pre-order: Financial Applications using Excel Add-in Development in C/C++, 2nd Edition

Publishing July 13th, 2007

Our quantitative finance team is involved in many parts of the banks business. We are not specialised to work in certain areas and markets like most quant teams in the city. We have to deal with a number of issues like portfolio optimisation, dynamic portfolio trading strategies or derivatives pricing. Furthermore, we have to consider the broad range of asset classes.

Because there is a limited amount of manpower available we look for methods that can be applied to a large class of problems. To this end we decided to implement some robust, flexible and widely applicable tools. One of those tools is the Monte Carlo method.

The Monte Carlo method can be applied to a wide range of problems in finance. We use it for examining dynamic trading strategies like CPPI, computing VaR and derivatives pricing.
Often Monte Carlo simulation is the only chance to get results. This is the case if you work on high dimensional, path-dependent problems. Monte Carlo simulation uses ideas from probability theory as well as calculus but also from number theory. There are many sources of input for new ideas and new methods. It is a very active field for researchers.

The method has – of course - some drawbacks. Some of the main criticisms put forward are that the procedure is time consuming when implemented on a computer and it is often hard to compute the Greeks and prices for American / Bermudan options. But latest research shows that difficulties can be overcome in applications. For example Giles and Glasserman developed new methods for computing Greeks or Schoenmakers tackled the problem for computing prices for Bermudan options in high dimensional Libor Market models.

Together with Daniel Duffy we work on a generic and robust method to implement the Monte Carlo method such that it can easily be extended and customised to fit your needs. This is hard work but we hope to come up with a nice setup.

The method of choice is programming in C++ using Microsoft Excel as a frontend. Excel is used for supplying data, outputting data and graphical display it. All the numerical stuff is done in C++. This gives us the opportunity to work object oriented and get speedy methods.

The complexity of the models used for derivatives pricing and portfolio optimisation does certainly increase. Complex stochastic models are put forward to capture the features observed in the markets. The Gaussian paradigm does not hold any more!

Therefore: Monte Carlo now! Not later!

This is my answer to the initially posed question. I suggest to set up a robust and flexible Monte Carlo engine. Then, you are able to tackle a lot of problems in finance appropriately.

Any comments?!

Cheers, Joerg