Harindra de Silva, who invests $59 million of other people's money in the UAM Analytic Enhanced Equity Fund, buys about 120 stocks a year. He doesn't talk to chief executives, visit the companies or listen to analysts.

Instead, the 39-year-old co-manager of the fund relies on a computer program to forecast stocks' one-month return based on 70 variables, including the ratio of companies' cash flow to share price, currency exposure and price-to-earnings ratios.

"I'm buying characteristics, not companies," said de Silva, a native of Sri Lanka who joined the fund in October 1996. "We haven't figured out a way to quantify the quality of management."

De Silva is one of the growing number of investors relying on computer models to help beat the benchmark Standard & Poor's 500-stock index. His strategy appears to be working. The fund, rated five stars by fund-tracking company Morningstar Inc., has gained 32 percent a year the past three years. It's up 9.2 percent this year, compared with a 4.3 percent gain for the S&P 500. Its biggest holdings are General Electric Co. and Wal-Mart Stores Inc.

Investors have flocked to the fund; it's swelled more than eightfold from just $7 million in 1997.

"There's definitely a trend toward using quantitative technology in fund management," he said. "The trend in quantitative funds is to focus on multiple characteristics instead of just one or two."

Overall, the success of such quantitative funds is mixed. Fidelity's quantitative offering, the TechnoQuant Growth Fund, for instance, trailed the S&P last year. It's up by 7.3 percent this year, beating the benchmark by 3.0 percent.

While many money managers use computer models to help pick stocks, few are as reliant on a machine as de Silva. He said stock-picking techniques that involve subjective approaches such as meeting management are irrelevant because they're "already reflected in the statistics." The key is to spend time determining which indicators are crucial to gauging investor sentiment, he said. He rebalances his variables during the first week of every month and has only broken with this schedule once--when the Russian stock market crashed, along with the ruble, in mid-July 1998.

Only four or five variables have a significant impact on the market in a given month, de Silva said, and they can change weekly. De Silva programs his computer to favor those variables when picking stocks.

For example, de Silva's computer told him Coca-Cola Co. would be a good buy one recent week because it sells most of its soft drinks overseas and the weaker dollar would make them less expensive. The beverage maker actually fell 5 percent that week.

"The benefit of this model is that we can look at many aspects of a company," he said. "Coke has a lot of things that will help it if the dollar weakens but many things that will hurt it in terms of the business model."

The computer also told him Circuit City Group might fall because it buys products with imported semiconductors that will become more expensive if the dollar sinks against foreign currencies. Circuit City fell almost 9 percent.

A ratio of cash flow to stock price was another indicator that de Silva's computer emphasized. Cash flow measures company performance by excluding non-cash charges and other expenses. Some investors who are skeptical of companies' efforts to manipulate earnings figures trust the cash-flow ratio more than the price-earnings ratio.

Cash flow, on the other hand, "is hard to manage--it's looked at systematically," de Silva said. "Cash flow and stock price returns have been highly correlated over the past six months."

De Silva developed his theories while working on his doctorate in finance at the University of California at Irvine. Based on the computer program, which he tweaks to accommodate new variables, de Silva buys and sells stocks once a month.

UAM Analytic is unique among quantitative funds because its turnover is high--about 300 percent annually. "It's doing what a lot of quantitative funds do," said Kinnel, a senior fund analyst at Morningstar. "But it's special because of that turnover."

Two of de Silva's top holdings, GE and Microsoft Corp., are the most heavily weighted stocks in the S&P 500. The high turnover rate, though, means it's expensive for investors who aren't in tax-sheltered accounts.

"It's a decent fund, but you wouldn't want to be taxable, given that environment," Kinnel said.

GE, the second-largest U.S. company by market value, makes up 4.25 percent of the total holdings of the fund. Retailer Wal-Mart accounts for 2.86 percent, followed by Citigroup Inc., 2.71 percent.

Technology companies make up almost one-third of de Silva's stocks. Microsoft and Lucent Technologies Inc., each make up 2.4 percent of the fund.

Transportation companies such as Delta Air Lines Inc. form the smallest industry group, with just 1.8 percent of the fund. Industry weights are correlated with S&P industry weights, de Silva said.