Steven Cohen’s Crowdsourced Quant Experiment Gets Real

Quantitative trading just took a step forward into its decentralized future. Quantopian, a start-up company that crowdsources trading algorithms and has backing from Steve Cohen’s Point72 Asset Management, has put investment capital into its first crop of algorithms.

The 15 algorithms to be funded, for an investment commitment in the range of tens of millions of dollars, were all developed by computer programmers using the Boston-based Quantopian’s proprietary research platform. The capital comes from an investment Cohen made last year, when he committed as much as $250 million in capital via Point72 to fund promising algorithms on the Quantopian platform. Point72’s venture arm also made a $4 million general partner commitment to the firm. Other Quantopian venture capital investors include Andreessen Horowitz and Bessemer Venture Partners.

Founded in 2013 by John (Fawce) Fawcett, Quantopian uses crowdsourcing and open learning to give the opportunity to build quantitative algorithms — computer programs that use data to identify tradable, and profitable, inefficiencies in the marketplace — to anyone with an interest in doing so. The platform, which includes education and training in how to build algorithms in addition to hosting competitions, events and an annual conference, has attracted would-be quants from all over the world.

The majority of Quantopian’s 120,000 members are not professional hedge fund managers or traders; most have computer programing or science backgrounds and are either interested in building algorithms as a hobby or as a side project as a way to break into finance. The algorithms now getting funded — with capital commitments ranging from $100,000 to $3 million — come from traders living on five continents in eight different countries, including Australia, Canada, China, Columbia, India, Spain and the U.S. When it finds an algorithm that proves to be profitable, Quantopian licenses the algorithm and pays the developer 10 percent of the profits; the developer keeps the intellectual property.

Fawcett says one thing he is pleased about with this first group of viable strategies is how much of the platform was used to develop these programs.

“The thing we are very, very happy with about the algos we are seeing is how they are using the full scope of the platform,” says Fawcett. The people building successful algos “were spending a lot of time on research.”

Quantopian CIO Jonathan Larkin says the increasing sophistication of the Quantopian offering is allowing the developers to build better, and more complex, strategies. “The algos themselves are becoming more complex, using a larger universe of names, and increasingly have a more nuanced aspect to them,” he says.

One of the key factors that makes crowdsourcing new, different, quant talent so intriguing to the likes of Point 72 is that it provides a way to avoid the problem of alpha decay — or the phenomenon in which any source of alpha or outsize return diminishes over time as the market becomes more efficient — at least in the short term. Fawcett and Larkin acknowledge that their first generation of algorithms will not last forever. Instead, they expect these strategies to last between three and five years (the normal lifespan for a mid-frequency trading strategy.) That’s why Quantopian is already looking to the crowd and testing its next crop of would-be trading algos.