QuantConnect Creates Tool to Develop, Share Algorithms

The Algorithm Framework, which provides a defined structure for developing algorithms, breaks algorithms down into five core features: universe selection (for asset class), alpha creation (for signal), portfolio construction (for how much to buy), risk management (for risk parameters and tolerance levels), and execution (for sending the order to the market).

Jared Broad, CEO of QuantConnect, which is an open-source, cloud-based algorithmic trading platform, says that the offering aims to serve as a foundation for quants to share and plug-and-play new models.

“The beautiful thing about the Framework is that we’ve got over 50,000 quants and they’re not all great at everything—some are really great at execution, some are really great at universe selection. So I can make an execution algorithm and I can give it to you and it will just plug into your algorithm,” he says. “Previously, you would have to read it, understand it, go through the code and figure out ways to make it connect together. It can be an ordeal. We hope that this will be the introduction of infrastructure, of architecture, to the quant world.”

A fund manager at DropShot Capital Management, which is a boutique hedge fund that specializes in artificial intelligence and machine learning—and which describes its focus as “algorithmic strategies without human bias“—tells WatersTechnology that the Framework will hopefully allow the firm to expand its offering.

“We use LEAN [QuantConnect’s algorithmic trading engine] extensively in our research and testing,” says the fund manager. “The new Algorithm Framework enables us to incorporate all the stages of algorithm development and deployment in a seamless fashion.”

More Signals

QuantConnect—which has over 1.2 million algorithms on the platform—spent about six months developing the Algorithm Framework, gaining feedback from the community on design and functionality. Broad describes the tool as something similar to the invention of power plugs and sockets.

“In your home, you’ve got [plug sockets] in the wall and you have power running through your home, so you just plug your laptop in and we don’t think anything of it,” he says. “But before they invented that standard for the plug and for the wall socket, you’d have to wire your laptop cable to the wall; you’d have to strip the cable, turn off the power and wire the copper wire together—that’s the common thing that’s done today in the quant world. [The Algorithm Framework] defines that interface—defines the plug—where one piece of code can fit into another.”

The vendor spent those six months of development making it more flexible since quants have very different ideas, and the Framework needed to be able to handle any sort of notion that a quant would look to throw at it, Broad says.

As a result, QuantConnect had to retool the Framework to take in different signals, Broad says. Take, for example, the famed Black–Litterman model, which was published by Goldman Sachs traders Fischer Black and Robert Litterman in 1992 and takes in several signals across various asset classes to compute the desired asset allocation.

“The Framework was previously only designed for ‘my’ algorithm, ‘my’ signal; it’s going into ‘my’ portfolio construction system,” Broad says. “So we had to rejigger it to accept multiple ideas and multiple inputs.”

New Marketplace

In addition to the Framework, which is fully live—the company is building tutorials for how to use it—QuantConnect is also close to going fully live with another new offering. In December, QuantConnect launched Alpha Streams, a marketplace for the QuantConnect community to list their alpha-generating insights for licensing.

Hedge funds and prop-trading shops can subscribe to the Alpha Streams API to receive thousands of real-time signals as a way to “crowdsource ideas,” Broad says, adding that QuantConnect is not a hedge fund, thus differentiating it from other so-called crowdsourced hedge funds in the industry.

Broad says that marketplace will hopefully be set to go in about two-to-three months, at which point, hedge funds can pick out strategies that they want and license their own to the community.