New York: Every night for a year, it was the same thing.
In a cramped, single-bed dorm room along India’s eastern coast, Ayush Gangwar put away his engineering books, cranked up the Bollywood show tunes and started typing out messages in the language of his real love, quantitative finance.
His best work came in the hours after midnight as silence enveloped the Indian Institute of Technology, ideal for poring over lines of code.
The 22-year-old logged onto a website connecting him to his boss in Greenwich, Connecticut, and began tinkering with models for souping up strategies at one of the world’s largest quant firms.
WorldQuant, a quantitative hedge fund that manages more than $4.5 billion for Millennium Management LLC, already has more than 120 full-time PhDs scouring everywhere for recurring patterns that might boost returns. But in this increasingly competitive corner of finance, that’s not enough.
So the outfit became one of the first to sponsor coding competitions and dangle the promise of a part-time gig in front of thousands of programmers like Gangwar to see if one of them might be the next James Simons or David Shaw.
Gangwar has everything WorldQuant could want—extreme computer literacy, enough hardware to dial in from 8,000 miles away, ambition—without a Wall Street professional’s demands on pay.
That’s not to say there wasn’t a steep learning curve.
“It was hard at the start,” said Gangwar, whose coding skills over two months of competing in the WorldQuant Challenge were good enough to land him a part-time job. “I had little knowledge in finance, so I had to Google a lot to learn the different terminologies.”
Gangwar, who entered the competition in 2014 and ranked in the top 50 among more than 5,000 contestants, says he earned roughly $7,000 in the year he worked as a WorldQuant consultant.
Crowdsourcing part-time quants as a way to find workers has become a full-blown craze on Wall Street in the past year. As active managers fight chronic underperformance and revulsion over their fees, funds are looking for ways to cut costs when it comes to talent.
Big names like Two Sigma Investments LP and startups like Quantopian Inc. are betting they can crowdsource geniuses to get an edge.
The method annoys sceptics who wonder if relying on the collective consciousness of data geeks is any better than hoping monkeys can write “War & Peace.”
But the criticism doesn’t faze Igor Tulchinsky, founder and chief executive officer of WorldQuant, who believes that with enough brains and technology, his firm can become king of the data processors.
“I’ve never seen a better time for asset management than today,” Tulchinsky said on a December panel during the Milken Institute London Summit. “There’s so much data out there, there are so many new technologies. It’s really a very exciting time full of opportunity.”
To Tuchinsky and a host of other funds pursuing the Darwinian approach to numbers crunching, the opportunity lies in volume.
Crowd-sourced quant is a radically different science than those practised even a decade ago by the founding giants of program trading, firms like AQR Capital Management that tried to pair known market anomalies with industrial-strength execution and beat the market over time.
Tulchinsky’s strategies, wrapped in secrecy, spring from a different proposition. It’s that if you have enough people examining enough relationships in the world, you’ll continually find ones that lead to market-beating returns. A deeper and more diverse talent pool also increases the odds that new insights could emerge.
On challenge number one, WorldQuant claims to looks at thousands of new information sources a year, no matter how exotic.
From those, it’s built a library of 4 million “alphas,” or pieces of predictive code that tell the computer to buy or sell.
Some may be simple, others may be attempts to take advantage of market anomalies. Portfolio managers then construct strategies by using the alphas as building blocks, depending on which the current market environment favours, swapping out ones that may have lost their edge.
The firm hopes to grow the library to 10 million alphas by the end of this year, Tulchinsky said in December, requiring the work of ever more quants—easier said than done.
Intense competition between funds for new hires from top schools and the infrastructure for hundreds of full-time quants makes it a costly endeavour. And talent may not live a commute’s distance from a WorldQuant office.
Hence the WorldQuant Challenge, in which the firm harvests brains of everyone from students in Boston, video game programmers in Russia, and even farmers in Taiwan.
Out of the 7,000 active users on their online platform, called WebSim, they’ve hired over 450 as research consultants. They have contributed 10,000 signals to the library.
Virtual workers allow for quicker growth versus exclusively hiring full-time employees, said Jeffrey Scott, director of WorldQuant’s Virtual Research Center, which heads up the Challenge.
To be sure, the Challenge is just one piece of the puzzle, and the competition’s main benefit is finding untapped talent, Scott said.
The cheapest labour comes from the firm’s machine learning programs, which have contributed a large percentage of the 4 million alphas, according to WorldQuant.
The consultants are responsible for their own equipment, according to a contract obtained by Bloomberg. There’s no clock for them to punch—the pay isn’t based on an hourly wage.
Instead, it’s a stipend paid out on anywhere from a monthly to yearly basis, along with a percentage of the returns from strategies built using the contractor’s alphas.
The pay and frequency of payment vary by country. Not everyone is in love with the idea of the crowd. Some believe it encourages data snooping, where participants set up complicated parameters to make the strategy look great in the past, while replicating it in the wild is another question.
“You’re not going to publish something that did not work in the past, and that doesn’t mean it’s going to work in the future,” said Rob Arnott, founder of investment adviser Research Affiliates LLC. “When you’ve got so many people looking at the same data as you are, the likelihood you find something new is pretty slim.”
But WorldQuant maintains it’s different. Existing infrastructure allows the company to dispassionately evaluate the programs contestants develop.
They apply the same rigor to evaluating potential hires as they do their own full-time data scientists.
For example, to avoid users gaming the system, WorldQuant runs tests to make sure that each alpha submitted is unique, and not just variations of old ones.
“What makes WorldQuant different is the technology we use, the people we have and the large quantities of diverse data we have access to,” said Scott. “Having so many part time consultants that are contributing unique ideas provides us greater opportunity to significantly diversify.”
As for Gangwar, he’s ditched his mining engineering major to focus on quant. He’s still in school but now has a different part-time gig. After his one year contract ended, Gangwar got together with two other friends and started to build the infrastructure for their own quant hedge fund to trade Indian equities.
Part of the seed money, said Gangwar, came from his WorldQuant consultant pay. Bloomberg