© 1999 Waters Information Services, Inc. All rights reserved. Used by permission. This is a reprint originally appearing in the January 22, 1996 issue of Trading Systems Technology. For subscription information call 607-770-9242.

A team of traders and technologists within Deutsche Bank North America's fixed-income group is evaluating a pricing, analytics and application development platform comprising Applix Inc.'s Applixware real-time application suite integrated with Tech Hackers Inc.'s @nalyst spreadsheet add-in library, the Mathworks Inc.'s Matlab development environment and Fame Information Services Inc.'s time-series database system. The DBNA team is now distributing prototype hedging and trading applications to front-office bond staff at the bank's recently constructed 324-position trading floor in midtown Manhattan.

"We're going to try and use software tools rather than native programming to meet our proprietary application development needs," says Stuart Sugarman, DBNA's head of technology. "If this [Applix-Matlab platform project] works, we'll avoid having to bring a significant number of programmers on board. Instead we'll rely on quantitative analysts and researchers that can use these tools to describe financial markets and prescribe trading models."

The task of designing and developing a trading analytics and application development environment for DBNA's fixed-income group was assigned last year to Mark Freed, a government securities arbitrage trader. "I'm an end user, but I also have some strong views on how to develop technology appropriate for the trading room," says Freed.

Freed adds that he had previously developed tools for his own decision support and trading needs. But these proved to be more generally applicable across DBNA's trading floor. "I was able to persuade the systems department that it was worth giving it a go," he says.

According to Freed, he first came across Matlab as a doctoral candidate in New York University's economics program. When he joined DBNA around a year and a half ago, he purchased a Matlab user license and began using its programming language to develop proprietary fixed-income arbitrage trading models.

With traders and salespeople across its new trading floor already outfitted with Applix's Unix-based, real-time spreadsheet, TechHackers' @nalyst series of financial function spreadsheet add-ins and the Fame database, Freed approached Mathworks' support staff with the idea of integrating the products to create a unified financial analytics and application development environment.

Once the interfaces between these products are fully tested, users will be able to import real-time market data into an Applix spreadsheet, run it through TechHackers' pricing/analysis routines and then transmit that data through proprietary C modules developed using Matlab. This information may then be displayed in two, three or four-dimensional Matlab graphs and updated on a real-time basis.

Wait and see
DBNA's project team and vendor support staff are enthusiastic about the results they've achieved thus far. But Sugarman cautions that "we won't know the full extent of Matlab's power--how easy is it to program with, how good is the interface to Applix, et cetera--until the analysis is complete."

End users within DBNA's fixed-income group access the Applix-Matlab platform via IBM and Sun Microsystems Inc. desktop workstations running IBM's AIX and Sun's Solaris operating systems.

The Applix-Matlab system is linked with Deutsche Bank's proprietary Market Data Infonet (MDI) digital data distribution platform (TST, Oct. 3, 1994) for accessing, storing and publishing real-time market data. MDI is the result of a collaboration between technology staff from DBNA and Reuters, whose Triarch 2000 system forms the core of the MDI platform (TST, May 4, 1992).

Each component of DBNA's platform has its relative strengths, say bank officials. Applix's real-time spreadsheet and TechHackers' @nalyst library of financial functions provides DBNA with links to real-time market data sources, relational databases and a consistent set of industry-standard financial calculations. Fame's database system allows traders to access, analyze and distribute large time-series datasets. DBNA technology staff are also working on linking the Applix-Matlab platform to the bank's Oracle Corp. relational database system, adds Freed.

The Mathworks' Matlab forms the core mathematical and statistical engine behind DBNA's new analytics platform. The product comes equipped with its own Fortran-based fourth generation programming language, an automated C-code generator and a C compiler.

Freed says Matlab allows trading staff to develop applications without getting embroiled in the intricacies of writing C code. "I believe we're getting results that are at least 90 percent as good as you'd get using [native] C programming with significantly lower costs," he comments.

The DBNA team is also making use of Matlab's Optimization Toolbox. One of a number of optional Matlab modules, the Optimization Toolbox provides users with nonlinear function optimization algorithms that can be used to design hedging strategies.

Execution and Development
The Matlab environment has been designed to balance speed of execution and speed of development, according to Dave Eiler, the Mathworks' manager of financial product development. Matlab's design and language enables end users to switch between editing programs and viewing their results in graphical or text-based formats. Such functionality makes the development and debugging process "more natural and quicker," says Eiler.

As an example, Eiler says Matlab's partial differentiation routines can be used by traders to create three- and four-dimensional surface plots commonly used in options sensitivity analysis with a small set of Matlab commands.

While Matlab's interpreted language is designed specifically to perform analytic computations quickly and efficiently, sooner or later a point is reached where compiled C code is going to be considerably faster, Eiler notes. "As interpreted languages go, Matlab is extremely fast, but there's an upper bound as to how quick you can get with any interpreted language."

If greater processing speed is required, developers can compile the C-coded programs automatically generated within Matlab by making use of its recently introduced C compiler. Users can then run the module in native mode at machine speed from within or outside of the Matlab environment, he says.

DBNA isn't the only bank to use Applix's Unix spreadsheets in conjunction with other systems to support its pricing and analysis needs. First Union National Bank of North Carolina has deployed a similar platform based on a combination of Applix and Infinity Financial Technology Inc.'s Fin++ financial object library. Other financial institutions using Matlab as an application development environment for derivatives pricing include Daiwa Europe in London (Derivatives Engineering & Technology, June 12, 1995) and Fuji Capital Markets in New York (DE&T, May 30, 1994).

Click here for the latest news headlines from Waters.