BOSTON (Reuters) - In the secretive world of hedge funds, algorithms are not shared because they provide the juice behind market-beating returns, and are a key reason why hedge funds charge their clients “two and twenty” - an annual fee equivalent to 2 percent of assets, plus 20 percent of gains.
Now startup company Quantopian offers a tantalizing proposition for software and financial geeks who want to trade like a hedge fund manager - but don’t want to pay those steep fees. The Boston-based firm is bringing together a community of people who build algorithms used for trading stocks.
Nearly 30,000 algorithms have been created from the Quantopian community. A few hundred have been made available for free on the firm’s website (www.quantopian.com).
With help from Quantopian executives, I created a simple algorithm that generated $502,900 in paper trading profits from a $1.07 million bet on bank stocks. The 47.3 percent return over two years (April 29, 2011 to June 24, 2013) beat the benchmark S&P 500 index by 31.6 percentage points.
Quantopian’s top-shared algorithms include one that uses Google search terms to predict market movements along with a tech stock momentum play that enters the market when prices are moving up quickly and exits when they drop quickly
Perhaps Quantopian’s most useful feature is a back-testing function that allows investors to see how their algorithms would have worked in the past. You also can do a dry run by plugging your algorithm into live stock data. Since June 25th, on that basis, my algorithm returned 1.4 percent vs. 2.4 percent on the S&P 500 benchmark.
“It appeals to the engineer in me,” Quantopian Chief Executive and Founder John Fawcett said. “Before you jump into an investment, you can do a lot of preparation and see what will happen.”
Fawcett, who earned an engineering degree from Harvard University, has experience turning hedge fund-oriented software into a big payday. Tamale Software Inc, a company he helped start, was sold to Advent Software Inc in 2008 for $70 million.
My investment strategy was a pretty basic buy-and-hold, value play. I spread $1.07 million (not real money) across a basket of about 36 bank stocks with market capitalizations between $100 million and $5 billion. I identified my stocks by using a screening function on Fidelity Investments’ website.
It took Quantopian about 20 minutes to write my algorithm in Python, a common programming language used by brokerages. Python carried out the automated instructions for my trade.
After loading the stock names and tickers on a spreadsheet, Quantopian plugged the data into an algorithm being shared on its website. We tweaked the instructions to implement my buy and hold strategy.
I bet $35,000 each on names like People’s United Financial Inc, a mid-size regional bank in Bridgeport, Connecticut and New Jersey’s Hudson City Bancorp Inc.
The theory was that these stocks were undervalued because they were trading at less than their book value (assets minus liabilities). The great thing about Quantopian’s back-testing feature is that it provides an immediate assessment of your trading smarts.
Though I was able to trounce the S&P 500 benchmark, only someone with an iron gut would have been able to stick with my strategy for two-plus years. For example, at one point in August 2011, I was down about 25 percent.
Then, during the three months that ended August 31, 2012, my strategy caught fire, with my basket of stocks rising 13.6 percent during that period.
The next step for Quantopian is to create a premium service that allows investors to plug their algorithms into the software interface of a brokerage for live trading. A pilot program is underway with Interactive Brokers.
Not to worry, though. The paper trading success has not gone to my head. I‘m sticking with my day job.
Reporting by Tim McLaughlin; Editing by Lauren Young and Phil Berlowitz