Mastering the Market

Friday, Apr. 26, 2002
In October 1998, the super-successful, super-secret hedge fund Long-Term Capital Management (LTCM) made too many over-leveraged bad bets on how bond markets would perform based on historic trends and collapsed to the tune of $4 billion, rolling the world's financial markets. Paul Donnelly, CEO of Ireland's Prediction Dynamics (www.predictiondynamics.com) muses that had his company's software, Crucible, existed back then, LTCM might have sidestepped its meltdown.

The startup company's software takes a quantitative approach to market analysis, using complex mathematical formulas and historical data to predict in which direction markets — or even individual stocks or currencies — are headed. It allows users to build and maintain models that presage market movements. Donnelly says the software relies on two key technologies: nonparmetric and multivariate. For the uninitiated, the former means that the software makes no underlying assumptions, even in its math. For instance, it doesn't assume — as did the formulas used by the errant LTCM traders — the infallibility of bell curves. While it doesn't precisely rely on chaos theory, Crucible takes into account that, well, the unpredictable happens. "Markets don't always behave in nice, disciplined ways — you have to allow for eccentricities," Donnelly says.

Of course, that doesn't mean that Crucible, or any systematic trading software, can anticipate something as bizarre and monumental as last September's terrorist attacks on the United States. Obviously some non-financial events are beyond prediction, even though they affect markets. But it can account for the possibility of, say, interventions of federal banks, or how the fall of a small Asian telecom might have a knock-on affect on Europe's telecommunication sector.

Multivariate technology means that markets and financial instruments react to many different contributing factors, and Prediction Dynamic's software relies on a large range of historical data. It's also capable of maintaining a huge number of models simultaneously — it can handle models based on 3,500 separate instruments at once.

The software is aimed at hedge funds. It allows managers to make more accurate risk assessments. But in the end, it's up to traders to decide whether to take those risks. It can also be used to spot and exploit opportunities in markets that are too subtle for humans to detect. Crucible may also find favor with risk advisors to fund managers. Funds are usually categorized by the amount of risk its investors wish to assume. And low-to-medium-risk funds sometimes inadvertently cross into higher-risk areas. Crucible can help managers avoid such mistakes.

CTO and founder John Carney is the brains behind Prediction Dynamics; he developed it as part of his doctoral research at Trinity College. Since its creation in June 2000 the company has received about $2 million in financing, and it launched its first product, Crucible, in February. Donnelly expects a strong revenue stream by next year, and first profits by 2003.

A number of well-known clients are now using Crucible in trials, but Donnelly says he is not yet at liberty to name them. That would risk jeopardizing those deals at this early stage, and risk-avoidance is something Prediction Dynamics understands all too well.