A Blog by Jonathan Low

 

May 23, 2012

Does Computerized Securities Trading Need a More Human Touch?

Twice in the last week, high profile embarrassments have engulfed smart, successful organizations. And we are not talking about the Lakers losing in the NBA playoffs or Sarkozy losing the French presidency.

The recent $3 billion unpleasantness at JPMorgan and the subsequent Facebook IPO cowflop (down 18% and counting) have both been attributed, at least in part, to computerized securities trading that failed to prevent - and may have exacerbated - the sort of nasty surprises everyone thought this economy was too sophisticated to have to endure in the modern era.

Why any sentient investor might entertain such thoughts is a mystery after the BATs disaster earlier this spring, the financial crisis meltdown in 2007 and, the 'grandaddy of them all,' the 1987 Long Term Capital Management fiasco which involved a sure-fire strategy using algorithms devised by several Nobel prize winners.

The challenge, as the Wall Street Journal, describes it, is that too many people with similar backgrounds, technology and information are chasing too few trades. The result, rather predictably, is that they fail to adequately take each others' existence into account, which is precisely what renders their models less than optimally beneficent.

The reality is that technology has not rendered us immune to undesirable results. In fact, it may have made us more susceptible to them because it has caused otherwise smart people to lower their guard. The solutions are neither painful, not particularly expensive. But they may require that most unpalatable of choices in a high testosterone world, an admission of that one might have been wrong. JL

Jenny Strasburg reports in the Wall Street Journal:
When Robert C. Jones started working on Wall Street 30 years ago, few investors placed their faith—and made big trading bets—in computer-driven models. The gospel eventually spread across the money-management industry, luring droves of programmers and physicists, and tens of billions of dollars, to banks and hedge funds. But the surge in so-called quantitative investing, followed by the scarring "quant quake" of 2007, left Mr. Jones, who had become a senior executive at Goldman Sachs Group Inc., reconsidering the role of machines.

"The reason I did quant is that I didn't trust human judgment, including my own," says Mr. Jones, who left Goldman in 2010 after 23 years at the securities firm. "Relating the data to the [trading] outcomes made a lot of sense, but now that's a tough, tough game, with too many people and too much access to the data."
The 55-year-old Mr. Jones now is changing his tune—somewhat. The answer to improving the computer-driven stock-trading model is to weave in research from analysts, but to leave out the biases and emotions that can creep into final trading decisions.

With that goal, he and fellow alumni from Goldman, BlackRock Inc. and elsewhere have launched an investment firm called System Two Advisors L.P. It is underpinned by the idea that quants must keep evolving in a market dominated by powerful computers able to mine ever-expanding sources of information.

System Two—which refers to analytical, rational decision-making, according to marketing documents reviewed by The Wall Street Journal—is starting with just $50 million. Mr. Jones and his team will break down stock research by a team of India-based analysts into clearly quantifiable rankings, a process they expect will capture the nuances of their judgments while eliminating the biases. They then plan to integrate individual stocks' scores into computer-driven models.

It is an idea the rest of the quant world is watching closely.

"Bob is an original quant pioneer," says Cliff Asness, another Goldman alum who worked with Mr. Jones more than a decade ago, and now oversees more than $50 billion in assets invested in quant strategies at AQR Capital Management.

Mr. Asness calls Mr. Jones's research "very intellectually interesting," but AQR doesn't do fundamental research and has no plans to use human analysts soon. "The cultures have a lot of difficulty combining," Mr. Asness says.

System Two's chief executive is Anupam Ghose, who most recently was a founding partner and co-president of Roc Capital Management, a Deutsche Bank AG spinoff that incorporates research by a team of analysts in India with a quant-trading system in New York. Roc launched with more than $1 billion, now has about $800 million, say people familiar with the firm.

The firms will be partners, in some respects: A big part of Mr. Jones's focus is to change the way Roc's 50 India-based analysts score thousands of companies on financial strength, growth prospects, valuation and other factors.

The goal: break down measures such as earnings potential and quality of management into quantifiable statistics. Mr. Jones and his team will take statistics generated by the analysts and plug them into forecasting algorithms, making the humans an integral part of the broader quant strategy.

Fundamental stock research is a Wall Street staple, of course. But Mr. Jones, a Grateful Dead fan and avid runner, views most of it as fundamentally flawed.

"The idea is to take the bias out of the humans," says Russ Wermers, an associate finance professor at the University of Maryland at College Park and adviser to System Two. Messrs. Wermers and Jones also are joining on a separate project, a quant-driven system for selecting mutual-fund managers called Arwen Advisors, after a "Lord of the Rings" character.

Quants like Mr. Jones suffered a bruising in 2007, when market shocks triggered computer-trading programs to dump millions of shares and other assets en masse before they reversed sharply and the market rebounded.

The episode showed how crowded the industry had become. "It was a shock," says Mr. Jones, who helped start Goldman Sachs Asset Management's computer-driven investing business in 1989.

As the 1990s progressed, he recalls, "We became overcrowded with talent. Resumes for these jobs had straight A's from MIT. You wondered, 'Why does everyone want to get into finance?'"

The computer whizzes wanted to snatch up profits from fleeting price moves in the markets, and their success attracted droves more like them. Meanwhile, their models developed troubling similarities, moving in unison in ripples of buying and selling when markets were stressed.

In 2007, bank proprietary-trading desks and hedge funds raced to sell positions during the quant quake. "You felt bad. The clients, all of these people, gave you money, and you wish you had seen it coming," Mr. Jones says.

Quant funds bled assets in 2007 and 2008, but have regained some of their footing since, investors and analysts say. Mark Carhart, another longtime Goldman quant-fund manager who left the bank in early 2009, set up an investment firm, Kepos Capital, which struggled for months to attract new investors, hovering at around $220 million. Since notching steady monthly profits, Kepos has attracted new clients, boosting the firm's assets to around $500 million, say people familiar with the matter. Mr. Carhart declined to comment.

Mr. Jones is more laid-back researcher than amped-up trader, favoring T-shirts and sports jackets for meetings. He never liked neckties while at Goldman, and now he really avoids them.

And like quant giant Ray Dalio of mega-hedge fund Bridgewater Associates, Mr. Jones is also a devoted practitioner of transcendental meditation. Now that he has his own firm, he'll promote "TM" among his employees, and fellow quants.

"'This would be good for you,' he told me. He knows I'm kind of tightly wound," says Mr. Asness, who's known for tantrums involving smashed computer screens, and who says he would consider meditation if it could be done competitively, like investing.

0 comments:

Post a Comment