When Hyde Park Global Investments began doing research into machine-based, automated high-frequency trading six years ago, "Many people were looking at us as if we were crazy," relates Adam Afshar, president of the Atlanta-based quantitative proprietary trading firm. Today computer-driven high-frequency trading is prevalent, he observes.

Like other machine-based, high-frequency trading firms, Hyde Park Global uses computers to detect short-term market inefficiencies or identify mispriced securities. "The challenge is how to capture these short-term inefficiencies that exist only for milliseconds," says Afshar, whose firm is in the process of colocating its servers in New York to achieve lower latencies (see related article). Hyde Park Global's trading platform has the capacity to execute up to 300 trades per second.

More and more firms, Afshar suggests, are turning to high-frequency trading because it offers a better opportunity for steady profits. "Everyone is doing this because it's very difficult to predict the markets," he explains.

But Hyde Park Global isn't the typical high-frequency trading shop: The firm has no traders, portfolio managers or analysts -- just physicists, computer scientists and mathematicians who do quantitative modeling, code writing and analysis, according Afshar. In fact the only two people in the firm with an M.B.A. are Afshar, who managed long/short portfolios mainly for offshore clients at Bear Stearns for 12 years, and Hyde Park Global's external chief risk officer, who works on finances and managing risk controls. The firm's five remaining employees have backgrounds in scientific fields, including one with experience in drug discovery, one who worked in nuclear physics and another with expertise in particle physics.

All of the financial analysis at Hyde Park Global is conducted via computers with machine-readable, publicly available information -- no human analysts are needed, Afshar insists. While the firm's primary experience is in equities arbitrage, it currently is developing an options model, he notes, adding that the firm also recently developed a pattern-recognition model with a computer scientist from Carnegie Mellon University that has proven extremely effective thus far.

Adaptive Models

Because no formula is going to work all the time, Hyde Park Global develops adaptive models, using a genetic algorithm (i.e, such as mutations and crossover), which are designed to respond to changing market conditions in real time, Afshar explains. While he refers to this as machine-based learning, he points out that the machines don't actually learn. Rather, "They recalibrate themselves within the parameters that you have identified," Afshar says, adding that they rely on data and quotes from previous trades to recalibrate.

Afshar uses the analogy of cruise control in an automobile to illustrate the adaptive capability. The car's computer controls the acceleration in response to changing road conditions in order to maintain the programmed speed. "If you go up the hill, it gives it more gas, and if you are going down the hill, it gives it less gas," says Afshar.

Similarly adaptive trading models analyze market conditions -- often executing dozens or even hundreds of calculations in milliseconds -- to identify mispriced securities based on preprogrammed parameters. "If you have programmed them correctly and given them the right calculations, and the analyses are correct, they can be extremely effective tools in identifying these short-term mispricings," says Afshar.