MIT Startup Flyberry Capital Emerges with Big-Data Hedge Fund
In the good old days—say, the early ‘90s—math and physics PhDs who wanted to make lots of money became quantitative analysts, or quants, on Wall Street. Now, they just start their own hedge funds.
It isn’t that simple, of course, but here’s a case in point: Flyberry Capital, a one-year-old startup in Cambridge, MA, founded by a team that includes MIT electrical engineering and computer science PhDs and an MIT Sloan MBA. Flyberry is using—can you believe it?—“big data” techniques to draw insights from huge streams of information, and applying them to finance. In particular, to inform trading decisions for its early stage hedge fund.
Flyberry manages what you might call a micro-hedge fund, soon to be on the order of $1 million, though it could be a lot more than that (see below). The firm uses its “global intelligence system” to gather information from hundreds of sources—such as sensor networks, weather patterns, earthquake detectors, government sites, news sites, blogs, and social media—and develops models of how markets react to various events. Then, based on those insights, the team makes trades in commodity futures, market futures, and other common types of trades.
“As much as we make use of a big data toolkit, we’re not a quant shop. We don’t do high-frequency trades,” says Zeid Barakat, Flyberry’s co-founder and chief strategy officer. “We don’t stay in the market, we just trade based on discrete events.”
If there’s a weather-related crop issue—like this year’s drought in the U.S. and its impact on corn—Flyberry tries to anticipate the market reaction. Or, with Hurricane Isaac bearing down on Louisiana, the firm could analyze years of historical data on storms (e.g., Katrina) and come up with some hypotheses. The same idea applies to any kind of event—say, the release of a consumer index report. Flyberry tries to deduce a “mathematical relationship between that event and the market,” says co-founder and CEO Michael Chang. “We gather a lot of information from the world, and try to test if this hypothesis is correct.”
Chang, who did his PhD at MIT in modeling and analytics, emphasizes that Flyberry’s technology platform is comprehensive. On the front end, it uses natural language processing to make sense of text-based data. On the back end, it’s optimized to test hundreds of hypotheses and develop multiple trading models so as to automatically generate a trading strategy. “It’s not a single algorithm or process,” Chang says. “It’s an end-to-end trade machine.” (As compared to most quant shops and funds, which focus on a few strategies, he says, “They have a fish. We have the boat and net.”)
And finance is just the beginning. If the team’s technology proves itself out over time, then it could also be used to help manufacturers manage their supply shops and predict plant failures, for example, or to create toolkits for risk management in investment funds, the firm says.
Of course, there’s plenty of skepticism to go around. … Next Page »