Institutional Investor's Alpha Magazine - March 2008 - (Page 26) Quant Investing corporate earnings. At the short end, investors’ immediate need to buy or sell can move a particular stock irrespective of the fundamentals. “The real question is, can such relationships be exploited?” Fludzinski says. Thales’s 35-person research team, which includes ten Ph.D.s, is looking for ways to forecast investor behavior on different timescales. (In total, the firm employs about 50 people.) Although Thales focuses most of its effort on improving current trading systems and strategies, Fludzinski says the firm devotes as much as 20 percent of its resources to finding a unifying theory of markets. Like physics, he says, finance lends itself to mathematical models based on observations. Through experiments that replicate what they see in the natural world, physicists have described its underlying principles. Thales is taking a similar approach with f inance, using computers to model socalled agent simulations of thousands of traders. In one experiment Thales created a — SANFORD GROSSMAN, CEO, simulation with 10,000 traders, QFS INVESTMENT MANAGEMENT each owning the same portfolio of 15 stocks. After hearing news about one of these stocks, half of the simulated traders might buy it and the other half might sell. Fludzinski says such computer models help explain why a stock typically trades higher than average for a few days after a positive earnings announcement and then trades lower before returning to normal. By modeling such herding behavior, he hopes to better understand how securities prices move in relation to one another. “Maybe we need to build a computer simulation that has 50 million people, with complicated rules for each,” Fludzinski says. “It’s very difficult to explain why people behave irrationally.” Fludzinski is quick to point out that Thales’s agent simulations are different from much of the work being done in behavioral finance, which applies cognitive psychology to analyzing markets by using models in which people don’t always act rationally, are prone to overconfidence and are more risk-averse when they are losing money. “This is not behavioral finance like you read about that’s kind of long term, where everyone doesn’t know their own objective function and the fear of regret is really high,” he explains. “That’s sort of macroscopic. What we’re doing is microscopic behavioral finance.” MIT’s Lo, whose adaptive markets hypothesis views the market as an evolving biological system, says behavioral finance has only recently started to gain credibility. “For the most part, the mainstream of financial theory has rejected psychology and behavior,” he says. “Even now they’re viewed with a great deal of skepticism.” Tanya Styblo Beder, chairwoman of New York consulting firm SBCC Group and a longtime quant, says behavioral finance is a vital aspect of any good trading operation. “I think it’s going to be huge,” says the former CEO of Tribeca Global Management, Citigroup’s now-defunct hedge fund division. “It will be one of the most critical ways for people to discover things about supply-demand flows in the marketplace. If you can figure out where the money’s going and what it’s coming out of, then you should be able to make trades and make some pretty decent dough.” Judith Posnikoff, a managing director at Irvine, California–based fund-of-hedge-funds fi rm Pacific Alternative Asset Management Co., thinks that Fludzinski and his team are well positioned for the current market. “They come up with interesting things on a regular basis and then actually implement them in the portfolio,” says Posnikoff, whose firm has invested in Thales since 1999. “And if something’s not working, they stop using it.” JOHN MOODY HAS KEPT HIS HAND in both finance and academia for more than two decades. Like Fludzinski, Moody, the founder of quantitative hedge fund firm J E Moody & Co. in Portland, Oregon, has a Ph.D. in theoretical physics from Princeton. He got into trading as a postdoctoral student at the Institute for Theoretical Physics at the University of California, Santa Barbara, in the early ’80s. From 1987 to 1992 the Portland native taught computer science at Yale University, where he was also a member of the neuroscience program. Moody then founded and directed the Computational Finance Program at Portland’s Oregon Graduate Institute. In the meantime, he had started consulting for clients ranging from New York– based J.P. Morgan Securities to the U.S. government’s Defense Advanced Research Projects Agency. The author of some 65 academic papers, Moody was among the first scientists to think about applying machine learning to finance. Machine learning is a branch of artificial intelligence whose proponents design computer programs that can recognize patterns and learn by trial and error. In the physical world, machine learning is used to help intelligent robots — for example, it enables the Mars rover to decide which of several routes to choose when moving through rough terrain. Financial markets, Moody says, are much less predictable and can be even tougher to navigate. “You’re trying to build programs that are able to identify relationships in data and draw inferences and make predictions from those relationships,” he explains. In 2003, Moody left full-time academic life to concentrate on his hedge fund firm. He also resigned from OGI to join the Algorithms Group, part of the International Computer Science Institute at the University of California, Berkeley. Moody, 50, has witnessed the financial world’s appetite for computer science, mathematics, physics, engineering and statistics graduates with knowledge of machine learning. He has hired several of his former students to work as traders and researchers at his firm and says “Instead of inventing things themselves, many firms are reading papers, talking with academics and coming up with strategies.” 26 • INSTITUTIONAL INVESTOR’S ALPHA • MARCH 2008
For optimal viewing of this digital publication, please enable JavaScript and then refresh the page. If you would like to try to load the digital publication without using Flash Player detection, please click here.