[投资心得] 福布斯：揭秘美国知名量化对冲基金Two Sigma

26934

收藏 2015-10-17

[相关阅读]
【华尔街系列】（资料汇总帖，附链接，持续添加中）

Rich Formula: Math And Computer Wizards Now Billionaires Thanks To Quant Trading Secrets
September 29, 2015

John Overdeck and David Siegel (right).jpg

LAST YEAR SOME OF THE nation’s greatest math minds gathered at Manhattan’s elegant Tribeca Rooftop for stunning views, fancy food and wine–and a series of complex equations to determine who among them would be king of the geeks. The annual event was hosted by the Museum of Mathematics, which was underwritten in 2012 by Google and the world’s richest quantitative Wall Street traders, including Renaissance Technologies’ billionaire founder, James Simons.

In the end there was little surprise that the finals featured Terence Tao, 40, considered one of the greatest mathematicians of this generation. A professor at UCLA, Tao has won the Fields Medal, often described as the Nobel Prize of math, and the USD 3 million Breakthrough Prize. But Tao didn’t win that night. He was beaten in the final round by a secretive Wall Street hedge fund manager named John Overdeck, who solved a problem involving infinite sequences and prime factorization. In fact, it was the second time in its short history Overdeck, also a backer of the museum, had won the competition.

Like Tao, Overdeck is a math genius, but unlike his contest rival, he never wanted to be a professor. After winning a silver medal at the International Mathematical Olympiad in Poland at age 16, young Overdeck told the Washington Post that writing papers didn’t seem fulfilling and instead he wanted to make the math “do something.”

Do something? How about build the fastest-growing big hedge fund on the planet and make multibillionaires out of its two founders? Overdeck, 45, and partner David Siegel, 54, run Two Sigma Investments, a little-known quantitative hedge fund firm that gathers seemingly random bits of information and tries to detect patterns that can be used to forecast the price direction of stocks and other securities. They are card-carrying members of the growing tribe of quants who use big data and machine learning in an attempt to beat the market consistently. And they join the ranks of the most successful of such traders, including James Simons (worth USD 14 billion), Ken Griffin (worth USD 7 billion) and David Shaw (worth USD 4.7 billion).

Within the last five years Overdeck and Siegel’s for-profit Wall Street think tank has quietly swelled from USD 5 billion to USD 28 billion in assets–one of the biggest hedge funds in America, even besting Simons’ Renaissance. What is yet more impressive is that Overdeck and Siegel’s rocket-science-fueled operation is able to command fees in its biggest fund of 3% of assets and 30% of profits, versus the industry standard of 2-and-20. The firm’s biggest fund, Spectrum, has earned an annual average return of 9.4% net of fees since 2004. Growth + returns + whopping fees = a great wealth-building formula. This year Overdeck and Siegel debut on The Forbes 400, each with a net worth estimated at USD 2.8 billion.

“The challenge I think facing the investment world is that the human mind has not become any better than it was 100 years ago, and it’s very hard for someone using traditional methods to juggle all the information of the global economy in their head,” Siegel said at an investor conference earlier this year. In fact, Two Sigma’s data scientists and systems analyze more than 10,000 data sources, using 75,000 CPUs with 750 terabytes of memory. Their hedge funds have executed more than 1.2 billion trades over the last 14 years.

Said Siegel: “Eventually the time will come that no human investment manager will be able to beat the computer.”

So in a world where investors should effectively be shorting the likes of Warren Buffett in favor of faceless machines like IBM’s Watson, where math formulas can create immense wealth, it is no surprise that Overdeck and Siegel are obsessive about avoiding publicity and keeping the firm’s secrets under wraps. (Overdeck and Siegel declined to comment to FORBES.)

But Two Sigma’s breakneck pace has created a big challenge for its founders. Putting all that money to work means stretching beyond its core competencies. The firm has expanded into reinsurance, venture capital and marketmaking. And it requires a never-ending supply of young math whizzes. Its core hedge fund operation, run out of offices in Manhattan’s SoHo district, now employs more than 800 researchers, computer programmers and statisticians, including 130 Ph.D.s and 6 International Math Olympiad winners. Most staffers are plucked straight out of the computer science, mathematics and engineering programs of MIT, Carnegie Mellon and Caltech. Instead of competing with Goldman Sachs and George Soros, Two Sigma opens its checkbook to compete for top talent with Silicon Valley firms like Google and Facebook.

But those big packages–a twentysomething researcher can take home USD 550,000–come at a price. Even in the sharp-elbowed trading culture, Two Sigma has proved exceptionally aggressive when it comes to protecting its methods and methodology, to the point where some employees who have tried to leave the nest have been sued, prosecuted–and jailed. These math nerds shoot to kill.

JOHN OVERDECK GREW UP IN THE well-to-do Maryland suburbs between Baltimore and Washington, D.C. His father was a mathematician at the National Security Agency, and his mother, who had a master’s in math, managed a computer company. At 16, Overdeck was studying at Stanford University, where he ultimately earned a degree in math and a master’s in statistics, but he dropped out before getting his doctorate. In 1992 he was recruited by D.E. Shaw, the quant hedge fund of billionaire computer scientist David Shaw. There he rose to managing director in charge of risk management but after nearly seven years left to work for another D.E. Shaw alum, Jeff Bezos, at his upstart Seattle-based online bookseller.

At Amazon, Overdeck was Bezos’ first shadow, following him everywhere he went. Then Overdeck was assigned to work on customer-relationship issues, building a services architecture for recommendations and customer reviews and supervising some 90 employees.

Rather than remain a well-paid cog in Bezos’ retailing juggernaut, Overdeck quit Amazon in 2001 to return to Wall Street to form his own data-driven hedge fund with Siegel, another D.E. Shaw refugee.

Siegel is the stereotypical computer nerd. A native of suburban Westchester County, N.Y., he is an ardent believer that technology makes everything better. He has a Ph.D. from MIT in computer science with a specialty in artificial intelligence. Prior to forming Two Sigma with Overdeck, Siegel worked at billionaire Paul Tudor Jones’ New York hedge fund, Tudor Investments. In fact, when Two Sigma launched in 2001–the early days of computer-driven “electronic trading’–Tudor Investments was the key investor, providing the startup with space in its offices at One Liberty Plaza.

What’s the significance of the name Two Sigma? According to those familiar with the firm, sigma with a lowercase “s’ is the ratio of an investment’s volatility to its excess return–that is, the return above benchmark. This is a key factor in deciding how much capital an optimal portfolio should allocate to a given investment. The second meaning of sigma–with a capital S–denotes sum.

From inception Two Sigma’s early funds, like Eclipse and Spectrum, focused on trading stocks globally. Eclipse was faster, changing positions within weeks, while Spectrum had a longer-term horizon closer to one month.

The duo eventually used their algorithms to create programs that operate outside the global stock markets, like the trend-following Compass funds that bet on futures markets. In 2014 another important fund, Horizon, was folded into Spectrum, which had diversified its offering beyond stocks. One of Two Sigma’s least visible funds is its Partners Fund, an internal fund of funds, fueled mostly by capital from the founders.

Inside the firm Overdeck presides over the construction of models while Siegel handles the engineering and infrastructure that support the technology used to make predictions. Think of Two Sigma as the kitchen at a top restaurant. Overdeck is the master chef, approving the recipes and preparing the meals. Siegel is the manager, ensuring that the kitchen is humming with ingredients, pots, ovens and electricity.

Two Sigma researchers spend time testing existing models, and each researcher is expected to come up with two or three new models per year. These are presented to Overdeck in a white paper that is typically less than ten pages long. Since Two Sigma’s trading models can change its forecast in seconds, lots of back-testing goes into each model. It’s not unlike the way Amazon exhaustively tests various Web-page changes in real time to ensure optimal clicks and purchases. At Two Sigma headquarters the model builders, who need to write code, sit with the engineers and collaborate with them all the time.

Two Sigma builds trading algorithms around four kinds of information: technical information like trading volumes of stocks; event-based information such as credit agency actions, mergers or other news; fundamental data like corporate financial statements; and so-called alpha capture, which is often company- or industry-specific intelligence, not publicly available per se and gathered via proprietary surveys. Alpha capture is controversial. In fact, Two Sigma suspended a survey of stock research analysts last year after BlackRock entered into a settlement agreement with New York’s attorney general, who claimed a similar BlackRock survey unfairly gave it access to information about companies ahead of other investment banking clients.

For many trades different kinds of models and data are used in combination. For example, the hedge fund may be signaled to buy a stock if a new analyst report is greeted with low volume. Cutting-edge trading tactics are also deployed, including the application of machine learning or artificial-intelligence techniques, where Two Sigma’s computers extract information, adapt to changing market environments and react on their own. When the humans overseeing the models intervene, it’s usually only to increase or dial back risk.

One of the biggest risks for Two Sigma is that its models work in theory but that the idiosyncrasies of limited data render them useless when applied to the real world. “It’s really hard to ascertain whether this kind of strategy has real merit or whether it’s a mirage of statistical computer might,’ says David Bailey, a mathematician and noted computer scientist affiliated with the University of California, Davis, who co-wrote a paper suggesting that these strategies are often “spuriously validated.’ Siegel emphasized this issue at an investment conference this year: “You have to formulate your big-data analysis in a way that you can understand whether or not you are over-fitting the data or actually extracting legitimate information out of the data–that is what the business is all about.’

While some big hedge fund investors, like USD 13 billion SkyBridge Capital, shun what they view as black-box quants like Two Sigma, there is no shortage of investors clamoring for consistent returns made possible by science. Since the financial crisis and the Madoff debacle, institutional investors have rushed to larger hedge funds with established systems and controls. By 2008 Two Sigma already had USD 4.6 billion in assets and 200 employees, while other similar-size value funds had far fewer employees.

Moreover, Two Sigma’s returns have been respectable and consistent. While many other big funds suffered losses this summer–the core definition of a “hedge’ fund has pathetically morphed from downside protection to the heads-I-win, tails-you-lose 2-and-20 payment model–Two Sigma deftly navigated the volatility. Spectrum was up by about 6% net of fees in the first eight months of 2015.

The USD 6 billion Compass fund had a big 2014, up 25.6%. Since inception in 2005 it has logged an annualized return of 14.9%. In August it gained 2.8% versus a 6% decline for the S&P 500. A leveraged version of Compass doubled that, up 5.7% during the turbulent month. Another big fund, Absolute Return, rose 6.1% in the first eight months of 2015.

Returns like these mean the money will continue to pour into Two Sigma. Last year, in fact, the firm raised USD 3.3 billion for a new macro fund, one of the biggest initial fundraisings in years.

扫码加我拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

全部回复

wwqqer

2015-10-17 20:40:19

TWO SIGMA’S OFFICES ARE THE KIND of place where you might trip over someone’s homemade remote-control car or look on as custom chess setsare created by 3-D printers in its hacker lab. On certain walls you can viewcolorful digital algorithmic art developed by an employee. There are hacknights, when staffers collaborate intensely on software-related projects like getting machines to play the video game Dr. Mario. Robots have been built to challenge staffers to games of shuffle board.

There are no suits or ties, just sweaters, collared shirts, khakis and jeans.People don’t work exceptionally hard by Wall Street standards, and they get paid well. Junior researchers in their 20s, for example, can make half amillion dollars a year in salary and bonuses. The firm has a retention rate of97%.

It sounds like an idyllic environment for would-be math professors or computer scientists–until they try to go off on their own. When Overdeck and Siegel feel threatened by the departure of a key employee, the lawsuits start flying, andon more than one occasion departing employees have faced criminal prosecution and jail time.

In 2006 Two Sigma sued Jianjun Qiu, a Two Sigma researcher who the firm claimed stole intellectual property and fled to his homeland of China. Two Sigma had hired Qiu in 2004, but within two years he was escorted out of its offices and placed on leave pending an investigation.

Two Sigma later claimed in court filings that Qiu had downloaded massive amounts of Two Sigma data to a home computer and refused to let the hedge fund inspect it. After being interrogated by the firm’s lawyers, Qiu sent an e-mailto his manager saying he had left for China. Two Sigma sent someone to meet with Qiu in China, where the hedge fund claimed he admitted taking “ten computerhard drives’ of data, which he refused to return. The hostilities ended abruptly after Two Sigma settled its lawsuit with Qiu, who returned to the U.S. and ultimately went to work at other financial firms, including billionaire Ken Griffin’s Citadel.

In 2010 an employee in Two Sigma’s Houston office came under suspicion and was summoned for a meeting at the Manhattan headquarters. Upon arrival he wasgreeted by New York City police officers and promptly arrested–Two Sigma suspected the employee had taken sensitive computer algorithms. The company alerted the Manhattan district attorney. The employee ended up pleading guiltyto a misdemeanor count of unauthorized use of a computer and left the firm.

Two Sigma’s most prominent criminal case against an employee involves Chinese national Kang Gao, a then 28-year-old researcher who gave notice to the hedgefund in 2014. Gao had worked at Two Sigma since 2010, after graduating from MITwith degrees in physics and engineering.

In early 2014, when Gao was preparing to leave, Two Sigma claims, he e-mailed himself some of the firm’s trading models. After exploring his options, heultimately accepted a position at Citadel but agreed not to start right away, in accordance with his one-year noncompete; he planned to spend his free time studying. Two Sigma notified Cyrus Vance Jr., the Manhattan district attorney,that it believed Gao had stolen its intellectual property. So after Gao’s recorded exit interview in February 2014, he was arrested on charges of unauthorizeduse of secret scientific material and unlawful duplication of computer-related material.

Gao pleaded not guilty but declined to meet the USD 500,000 bail set by the judge.Gao’s lawyer, Marc Agnifilo, reasoned that because Gao was a resident alien hewas headed to jail one way or the other. Two Sigma informed the government of Gao’s situation, and his work visa was revoked, depriving him of immigration status. He would thus have been transferred to federal immigration jail if he put up bail, leaving his lawyers with limited access to him as they mounted his defense.

Instead the boyish-looking, bespectacled Gao chose to apply for a tourist visaand wait for approval in the local jail, New York City’s notorious Rikers Island. During Gao’s eight-month detention he suffered through extended lockdown periods and was beaten up. While Gao’s IP-theft case never gained thenotoriety of that of Sergey Aleynikov, the former Goldman Sachs programmer sentto the slammer and immortalized in Michael Lewis’ book Flash Boys, Gao hasendured nearly as much jail time as Aleynikov, whose conviction has been overturned.

The criminal case against Gao accuses him of taking two trading models hedeveloped in 2011 and 2012. He was also accused of taking a research paper anda research presentation that he wrote. Says Gao’s lawyer, “How can taking mathbe secret scientific material? It’s math.’

In February 2015, just four months after leaving Rikers, Gao pleaded guilty toone count of unlawful duplication of computer-related material and wassentenced to time already served. He then returned to China. This hasn’t stopped Overdeck and Siegel in their quest to punish him. In addition toaccusing Gao of taking its property, Two Sigma filed a civil suit against Gaoaround the same time he was arrested.

The judge in the case, Jeffrey Oing, has been openly hostile to Overdeck and Siegel’s hedge fund. “Aren’t you guys going over the top in terms of this, toput someone in jail for a breach of duty?’ he said in a 2014 hearing. “What caused him to have you guys, other than being totally annoyed and angry at him,to do this?’ Now Two Sigma is initiating an arbitration case against Gao,seeking over USD 300,000 in damages. Gao’s lawyer has tried to stop it by arguingthat his client has already spent USD 100,000 on the civil action alone.

In early 2015 a senior analyst at Aberdeen Asset Management, an investor in TwoSigma, told Two Sigma’s London office that Aberdeen had been approached by aformer Two Sigma employee about a new hedge fund. Other firms like DeutscheBank and UBS were similarly approached.

The inquiry came from Sergey Fein, 34, a computer scientist recruited fromRussia’s St. Petersburg National Research University. Fein had worked at Two Sigma in Manhattan for seven years, ultimately rising to the level of vicepresident. Unlike most researchers at Two Sigma, Fein had worked on all four of Two Sigma’s proprietary investment disciplines. When he resigned in May 2014,Two Sigma elected to apply its one-year noncompete and pay him USD 210,000, 40% of his USD 525,000 annual compensation.

But in March 2015 Overdeck and Siegel got wind of Fein’s discussions about a new quant fund. Two Sigma then delivered a threatening letter to Fein’swould-be partners. Within days Fein’s plans fell apart as partners abandonedthe effort. But that wasn’t enough. The next month Two Sigma sued Fein in a New York State court. Fein has denied violating his contractual obligations to TwoSigma, saying he and his partners held only exploratory talks with potential investorsand didn’t raise any money or open a brokerage account. Fein said he hoped to raise the money after the noncompete expired.

“I have to ask Two Sigma a question,’ Judge Oing began during the lawsuit’sinitial hearing. “Did you seek to put Mr. Fein in jail?’ Oing continued: “It sounds to me like it’s like the same thing that happened with the otheraction.’

Two Sigma did not contact law enforcement but wanted the judge to extend Fein’s noncompete for seven months, the length of time by which Two Sigma claimed hehad breached the initial deal. Judge Oing refused, saying he didn’t have theauthority, and closed the case, ordering Fein not to violate the initial deal for its remaining five weeks.

Fein should consider himself lucky. Overdeck and Siegel have concocted arguablythe greatest hedge fund success story of post-meltdown America. And they’veclearly done the brass-knuckled math: When that success is derived around formulas and equations, you do whatever is necessary to keep your numbers to yourself.