• A group of young math and computer experts with roots in Michigan is using artificial intelligence to boost the performance of an $11-million investment fund they operate in New York.
The idea for tapping artificial intelligence to make better investment decisions than humans was developed by Spencer Greenberg, whose grandfather, Hank Greenberg, was an All-Star and MVP first baseman for the Detroit Tigers during the 1930s and 1940s.
“Google and Netflix use a branch of artificial intelligence called ‘machine learning’ to improve Web searches and predict what kinds of movies people will watch,” Greenberg says. “I had the idea to apply machine learning to the stock market.”
The computer system routinely crunches millions of pieces of research findings, including a decade or more of market data, and compares the information with up-to-date investment activity. Often, “Star,” as the program is dubbed, will select cheap stocks or those that rise swiftly in price. The company does not borrow money for its trades, and it doesn’t short stocks or use leverage.
Despite a few bumps, Greenberg, 27, says he and his three partners at Rebellion Research have largely outperformed the Dow Jones Industrial Average. Star reviews around 30 factors that affect a stock’s performance, such as interest rates. The company holds approximately 70 stocks at any given time.
Every weekday morning, Star offers up a list of stocks to buy, sell, or hold. While the firm’s partners — including Kalamazoo native Jeremy Newton, chief science officer — implement the trades, the team has yet to overrule the computer. So far, the artificial intelligence system has beat out more traditional indexes.
In 2007, Star gained 17 percent for the year — well ahead of the 6.4 percent advance of the Dow industrials. In 2008, the fund dropped 26 percent, due to the global financial meltdown and other issues. By comparison, the Dow was down 34 percent.
In 2009, Star had positive returns, up 41 percent, while the Dow rose 19 percent.
For 2010, Rebellion was up 21 percent, while the Dow finished with a gain of around 11 percent.
“Right now we’re working on a market-neutral version of our product so we have long and short products in roughly equal amounts,” says Newton, who helped give “life” to Greenberg’s computer model.