Hedge-fund managers are overwhelmed by data, and they're turning to an unlikely source: random people on the internet
Jim Davis/Getty Images
- With the amount of data available growing astronomically, hedge funds continue to search for the most efficient way to filter through it all to gain an edge, all while industry margins shrink as fees drop and outflows continue.
- Open-source platforms such as QuantConnect and Quantopian are now offering hedge funds data analysis on the cheap, either through investing in a fund or licensing a freelancer's work. Managers who have been notoriously secretive must now decide if they want to open up their processes to freelancers online.
- "Hedge funds are going to see that being closed isn't a competitive advantage anymore," Jared Broad, the CEO of QuantConnect, a platform with 75,000 engineers who dive through datasets looking for investment signals, told Business Insider.
- For more stories like this, visit Business Insider's homepage.
Hedge funds are sifting through so much data that they might just turn to random people online to help with it.
Alternative data streams of satellite images and cellphone-location data are where managers are now digging for alpha, as new datasets are created every day. And hedge funds have been spending serious cash searching for those who can take all this information and quickly find the important pieces.
Now, as margins shrink and returns are under the microscope, hedge funds are beginning to consider a cheaper, potentially more efficient way to crunch all this data: open-source platforms, where hundreds of thousands of people ranging from finance professionals to students, scientists, and developers worldwide scour datasets - and don't get paid unless they find something that a fund finds useful.
While funds might be desperate for help with the data, the risk of a manager's secret sauce slipping out to the masses has often stopped many from putting these platforms to work.
But open-source platforms like QuantConnect and Quantopian - which give users tools and datasets to create their own algorithms and find investment signals - are becoming increasingly popular as secretive hedge funds move toward a more collaborative approach.
Three funds - Tibra, Maritime Capital, and FME - have subscribed to license signals found by freelancers on QuantConnect, and billionaire Steve Cohen has backed Quantopian through Point72's venture arm.
Their appeal is like an insurance policy on a manager's data: If you can't find anything worthwhile in it, let thousands of others try to play around with it.
To be clear, these platforms don't require hedge funds to submit their entire portfolios or trading techniques in order to review coders' work. But to incorporate a signal or algorithm made by a freelancer, funds have to at least partially let an outsider into their systems and processes, which many have historically been reluctant to do.
That's starting to change.
Shayanne Gal/Business Insider
Citigroup and Goldman Sachs have opened some of their trading platform's code up to the public, and AQR, Two Sigma, and D.E. Shaw have let outside coders use their artificial-intelligence and data-processing tools.
"Hedge funds are going to see that being closed isn't a competitive advantage anymore," Jared Broad, the CEO of QuantConnect, a platform with 75,000 engineers who dive through datasets looking for investment signals, told Business Insider.
A surplus of data, and a shortage of data scientists
John Fawcett started his open-source platform, Quantopian, with a simple question in mind: What if there were more quants?
In 2011, when the platform was launched, Fawcett estimated there were between 5,000 and 6,000 professional quants working in investment management. The issue he saw, and believed his platform could fix, was the expected increase in data was not matched with a similar increase in quant hiring.
"There was just an incredible bottleneck of this ecosystem of data filtering into portfolios," he said.
Fawcett's platform has since expanded to 250,000 members around the world, and 4 million algorithms have been created, as coders, often with no finance background, get access to datasets and other tools that were historically closed off to those not at the biggest funds. The core of what QuantConnect and other platforms do, according to Broad, is "democratizing vast amounts of financial data."
Fawcett's Quantopian can license any of the algorithms for its hedge fund, which has $50 million in assets and money from Cohen, into community-created algorithms, paying the creators of these algorithms more than $300,000. To get exposure to the best algorithms generated on Quantopian, hedge funds can invest into Fawcett's fund like Cohen.
The view from Point72, Fawcett said, is that there is a scarcity of talent in finance.
"We're at the beginning of this explosion of data, and the bottleneck is the community of professional quants," Fawcett said. "If you're trying to find predictive data, then you really need to look at all of it."
Tibra, a quant fund that subscribes to QuantConnect's platform, is using the database of hundreds of thousands of investment signals to help make sure it never stops trying to improve, Tibra CEO Chris Udry said in a blog.
"One of the most challenging aspects of running a successful trading company is the continued evolution of its investment process," Udry said.
The ability for Udry's quants and coders to be able to incorporate community-created signals into the pre-existing algorithms will make QuantConnect and others "the platform of choice for quants," Udry said.
Balancing cost savings with security concerns
Protecting a fund's secret sauce is always going to be the biggest concern for any manager allowing its quants to collaborate on open platforms.
"[Hedge funds] haven't fully embraced those elements of collaboration and figured out how to keep proprietary stuff close to the vest while sharing the more boring stuff," Tosha Ellison, a director at the Fintech Open Source Foundation and a former executive at Credit Suisse, told Business Insider.
Even switching to operating systems on the cloud took longer than expected because of security concerns, said Two Sigma's head of technology Alfred Spector.
"We're acutely aware in the domain in which we operate that we have proprietary algorithms," he said in an interview with Business Insider.
But with managers set to spend billions of alternative datasets, there is pressure on the data scientists and quants to turn that investment into alpha, and more manpower is needed - and it isn't cheap.
"If you're a hedge-fund manager today, you have to hire a quant, so that's recruiting and bringing this person on, getting systems set up, and then maybe in six months or a year, they will have something for you," Broad said. Compared to licensing data-driven signals from QuantConnect, this path is a lot more expensive, Broad added.
Indeed, "increasing cost constraints" is driving more financial services firms to open source, Ellison said. Top funds battle not only with each other but also with Silicon Valley for people who can efficiently grind through large datasets and model them into something useful.
What some long-running quant shops, like $94 billion manager Acadian Asset Management, have found though is that these open-source platforms eliminate the simplest of signals, making the quant space more competitive. The investment opportunities that were, at one time, able to be found only by professional quants can now be uncovered by someone playing around on Quantopian, according to Jim Dufort, Acadian's director of investment analytics and data, forcing professional quants to increase the amount of data they mine and the money they spend.
"We need to be in a position to differentiate ourselves with our clients by advancing our research, data, and tools beyond what's offered by both our competitors as well as in open source," Dufort said. "The most basic signals can be, and largely have been, commoditized."