Who's on the other side of the trade?

A great conversation between Tyler Cowen and fund manager Cliff Asness, who has appeared many times on this blog. See, e.g., this 2004 post on his analysis of the well known Fed Model for equity valuation, also discussed in the interview.

Hedge-fund manager Cliff Asness, one of the most influential—and outspoken—financial thinkers, will join Tyler Cowen for a wide-ranging intellectual dialogue as part of the Conversations with Tyler series.

Asness is a founder, managing principal and chief investment officer at AQR Capital Management. In 2012, he was included in the 50 Most Influential list of Bloomberg Markets magazine. As an entrepreneur in the field of finance, Asness has helped shape the national conversation on financial markets and regulation.

He is an active researcher and has authored articles on a variety of financial topics for many publications, including The Journal of Portfolio Management, Financial Analysts Journal, and The Journal of Finance. Prior to cofounding AQR Capital Management, he was a managing director and director of quantitative research for the Asset Management Division of Goldman, Sachs & Co. He is on the editorial board of The Journal of Portfolio Management, the governing board of the Courant Institute of Mathematical Finance at NYU, the board of directors of the Q-Group and the board of the International Rescue Committee.

Alpha geeks

More nonlinear returns to brainpower.

This Times article, about Google in China, features Kai-Fu Lee, a CS PhD (speech recognition) who was the subject of a bidding war and lawsuits between Microsoft and Google. He now runs Google's Beijing office (he used to run Microsoft research there).

This WSJ piece profiles Goldman's Global Alpha quant hedge fund. Asness has appeared before on this blog (you can search on the right), both for his research on a modified Fed model for equity valuation, and via a Times magazine profile.

In early 1997, Mark Carhart was an academic at the University of Southern California. His big claim to fame was his doctorate work at the University of Chicago on mutual-fund performance.

Today, the 40-year-old Mr. Carhart and another former Chicago-school colleague run a big, secretive hedge fund at Goldman Sachs Group Inc. which, with an estimated $10 billion in assets, is the Cadillac of a fleet of alternative investments that have boosted the earnings at the blue-chip Wall Street firm. And the two men are making millions themselves.

Known as Global Alpha, the Goldman hedge fund was a leading contributor to a surge in "incentive fees," or performance-related fees, that Goldman reported for the first quarter ended in February. In that period, the incentive fees soared to $739 million from $131 million a year earlier, helping Goldman's earnings rise 64% to $2.48 billion, the biggest first-quarter gain of any major Wall Street firm.

Global Alpha's recent returns have been sizzling. In the 12-month period ended in March, the fund returned more than 48% before some fees, according to Goldman Sachs JBWere, an Australian affiliate of Goldman. It was closed to new investors last year. (On Wall Street, the word alpha refers to investment returns beyond those generated by the market.)

The bearded Mr. Carhart and his colleague, Ray Iwanowski, manage Global Alpha. A 50-member team they lead also offers a menu of services for Goldman clients based on statistical models first developed by a group led by Clifford Asness, another former Chicago student.

The Global Alpha fund was seeded in late 1995 with just $10 million, and in its first full year, 1996, the fund returned 140%, one former group member recalls. Mr. Asness left Goldman in 1997 with seven of the group's 13 members, to form his own hedge-fund business, AQR -- short for Applied Quantitative Research.

...The Global Alpha group's lineage traces to a group of students of Professor Eugene Fama, an influential Chicago finance professor known for a belief in efficient markets. In the early 1990s, his former teaching assistant, Mr. Asness, was recruited to join Goldman by a college friend, and in turn recruited numerous colleagues from Chicago.

One of the group's early assignments was to build quantitatively oriented asset-allocation models. Part of the methodology, which underpinned the strategy of Global Alpha, was to select stocks selling at cheap prices based on their book value, earnings or other metrics, while betting on a decline in stocks selling at higher prices based on their growth prospects.

The Goldman group later used similar methods to choose not only stocks but also bonds, currencies, and entire country markets, former group members say. The models also included a "momentum" factor based on which stocks or markets have recently performed well. Although the models have evolved, the underlying "quant" methodology remains similar.

In addition to Global Alpha, Goldman employees manage two other hedge funds specializing in quantitative stock and bond investments with an estimated $8 billion in assets. While some competitors grumble that Goldman traders could gain an advantage based on possible access to client information, the Goldman funds don't have such access, former group members say.

The Goldman "quants" also offer a product similar to the Global Alpha fund known as global tactical asset allocation, which gives pension funds and other institutions the chance to boost returns using statistical methods. Goldman's 2003 annual report featured a team including Mr. Carhart which invested $1 billion for the General Motors Corp. pension fund using "active alpha investing."

The Goldman team continues to rely on the latest work from academia. Maryland's Mr. Wermers recalled a visit to Goldman about a year ago to discuss a paper he had written on whether the flow of investments into top-performing mutual funds could predict whether stocks they held would rise in price.

When he presented his findings in a boardroom filled with about 20 GSAM employees, Mr. Wermers recalls, "it was kind of a high-pressure event. They asked very, very tough questions."

Portrait of a quant

Here is a nice profile (NYT Sunday magazine) of hedge fund manager Cliff Asness. Some time ago, we discussed his research on a modified Fed model for equity valuation. Asness has a PhD from Chicago, and a quantitative style of investing. The article does a good job of explaining how hedge funds caught on with university endowments as alternative investment classes with (potentially) low correlation to the overall market, and good risk-return characteristics. Schools like Yale and Harvard led the way, with spectacular results.

Of being super rich, Asness says "Well, it doesn't suck." For a description of how the Bush tax policies favor the super rich, see here. (Those earning more than $10 million a year now pay a smaller share of their income in taxes than those making $100,000 to $200,000. So much for progressive taxation!)

Fed model reconsidered

The Fed model for equity valuation compares the E/P of stocks to the 10yr yield on Treasurys. The usual justification is that stocks and bonds are competing asset classes, and one should compare their future cash flows to obtain a relative valuation. When yields on bonds are low, investors will tolerate a lower E/P (or higher P/E) in equities. One subtlety here is future inflation, which seems to "pass through" to corporate earnings, but erodes the real returns on bonds. While E/P might be a plausible forecast of future real corporate cashflows, the 10yr yield is only in nominal dollars. Perhaps it would be better to substitute the 10yr yield on TIPS for the bond component.

I found some interesting analysis of the Fed model (and the following figures) in this paper by C. Asness. Figure 2 shows that the Fed model has been quite successful over the last 30 years, but not for earlier periods. I had always thought this discrepancy was explained by inflation - the Fed model was successful in the recent period when inflation was perceived to be under control (i.e., post Volcker). Asness has a different take. He fits E/P = a + bY + c v_s / v_b where Y is the bond yield, v_s the trailing 20y stock volatility and v_b the trailing 20y bond volatility, reasoning that the relative perceived vols will affect the attractiveness of stocks vs bonds. The result, shown in Figure 4, is quite nice. The best fit value of b is close to 1 (similar to the Fed model), and since the trailing 20y vol is by definition slowly varying, it seems the Fed model is not a bad rule of thumb for current valuation.

Simons, Thorp and Shannon

Nice article on Jim Simons and Renaissance in Saturday's Times. Part of me wants to get into the new fund (fees are pretty reasonable compared to Medallion), but then again they are embarking on something new, so it's no sure thing. See earlier posts here and here. Most impressive about Medallion is their consistency -- no down years since 1988 and only a single down month in the last 5 years!

BTW, I've received my copy of Fortune's Formula and it's quite good. I learned a number of tidbits about Thorp (the mathematician who wrote Beat the Dealer and invented a system for counting cards in blackjack), Shannon (the father of information theory) and others from this book. Apparently, Thorp and Shannon's investment returns (Thorp ran an early hedge fund called Princeton-Newport, while Shannon invested his own account) rivalled those of the best managers like Buffet and Soros. Buffet and Thorp actually knew each other early on, and had very high opinions of each other. The stories of Thorp testing his card counting system in Nevada are hilarious -- the level of detail after all these years suggests a phenomenal memory!

The bit about optimizing geometric vs arithmetic returns (a subject of controversy between math/physics guys like Kelly, Shannon, Thorp and economists such as Samuelson, and the origin of the title of the book) seems not so interesting to me, as the answer depends on what one wants to achieve. On the subject of hedge funds, it appears everyone is starting one, including information theorist Thomas Cover and former physicist turned AI researcher Eric Baum (author of What is Thought?, the best book I've read on AI).

NYTimes: $100 Billion in the Hands of a Computer

By JOSEPH NOCERA
PEOPLE ask me all the time: What's your secret?" James Simons said. We were sitting in an office in Manhattan that Mr. Simons uses when he's not at the Long Island offices of Renaissance Technologies, the money management firm he founded in 1982. He was wearing an elegant shirt and tie, and loafers with no socks. He took a drag from a cigarette, the second of three he would smoke in the course of a long interview.

I had indeed come to ask him what his secret was. In the hedge fund world, that's what everybody wants to know.

Mr. Simons, 67, who rarely talks to journalists, is hardly a household name like Warren E. Buffett. But Mr. Simons, who got into the hedge fund business after abandoning a stellar career in mathematics, has a track record that is jaw-dropping. This summer, word leaked out that he was starting a new fund - people took to calling it the "$100 billion fund" because its marketing materials say that it could conceivably grow to that enormous size. Not surprisingly, that has caused Wall Street types to be even more curious about him.

Here are Mr. Simons's numbers: from 1990 to 2004, Renaissance's primary hedge fund, called Medallion, has delivered annualized returns of 33.21 percent. (The Standard & Poor's 500-stock index has returned, on average, 10.98 percent during those same years.) Since the end of 2002, the fund, which has $5 billion under management, has disbursed $4.9 billion to its investors - with another $1.5 billion to be delivered at the end of this year.

And these returns are after Medallion's 5 percent management fee and 44 percent share of the profits - surely the highest hedge fund fees in the land. Medallion's returns, and its fees, have helped make Mr. Simons a very wealthy man, with a net worth that Forbes estimates at $2.7 billion.

When I showed Mr. Simons's returns to a hedge fund friend, he looked startled. "Nobody has numbers like those," he said. But here's the real eye-opener: no one outside the firm's 200 or so employees has a clue how he does it.

Medallion, you see, is a quantitative fund. In quant funds, trading activity is generated by complex computer models rather than human judgment. Most quants are secretive about the algorithms that drive their models; after all, that's their investing edge. But of the handful of big-time "black box" investors, as they're often called, Mr. Simons's box may well be the blackest.

HERE'S what we do know. Medallion's portfolio contains literally thousands of stocks and other financial instruments that it trades in rapid-fire fashion. The firm's scientists are constantly searching for repeatable patterns, and other signals, in the enormous amounts of data they compile. The computer models they devise tell them when to make trades based on those signals.

As Mr. Simons put it - and this is about as specific as he would get - "Certain price patterns are nonrandom and will lead to a predictive effect." He also told me that Medallion sticks with highly liquid securities that trade in public markets around the world. Why? "Because there is a lot of data on such instruments, and we're very statistically oriented," he said. He stays away from exotic derivatives.

Not even Mr. Simons's investors know much more than I've just described. "We trust Jim and we think he's smart," said one longtime Medallion investor. "So we stopped caring what the computer was doing." When this investor began describing Mr. Simons's investing approach, he admitted he was guessing.

Mr. Simons shrugged when I suggested to him that his firm's lack of "transparency," as they say in the business, was bound to make people nervous. Humans fail in the market all the time, but somehow we are willing to keep giving our money to human beings to manage because we understand investing based on human judgment. Or at least we think we do. But black box investing feels different. It feels scary somehow, precisely because it is not something most of us can understand.

"How any great investor does it isn't in the least obvious," Mr. Simons responded. "How we do it isn't any more mysterious than how a great fundamental investor does it. In some ways it is less mysterious because what we do can be programmed." Then he stopped, took another drag from his cigarette, and let out a small chuckle. "Well," he conceded, "it's less mysterious to us."

Mr. Simons wasn't always a quant. A former crypt analyst - a code breaker, that is - he did important work in mathematics that helped lay the foundation for string theory. When he began managing money in the 1970's, he did it the same way most investors did: he used his own judgment. "At first," he said, "I didn't think about investing in a scientific fashion. But I was trading currencies, and it gradually occurred to me that there might be some way to create models that would allow you to predict currency movements."

Although Mr. Simons and a partner made an absolute killing in the currency markets the old-fashioned way - they made huge bets that turned out to be right - he began surrounding himself with scientists who developed models for all sorts of tradeable securities. "By the end of the 1980's," he said, "I was a model man, and didn't want to do fundamental analysis." One advantage, he said, is that "models can lower your risk." Another, though, is that "it reduces the daily aggravation." With old-fashioned stock picking, he said: "One day you feel like a hero. The next day you feel like a goat. Either way, most of the time it's just luck."

Indeed, trading the way he does, making thousands of small trades aimed at capturing small price movements, doesn't generate the kind of "10 bagger" that investors love. But when done well, quant investing is less likely to have the kind of disaster that is always the danger when one bets big on a stock.

To those who point to Long-Term Capital Management as an example of the dangers of black box investing, Mr. Simons's defenders point out that his fund has far less leverage than Long-Term Capital, and that in any case, while Long-Term Capital had several Nobel laureates on board, human bets were what caused it to go awry.

Clifford Asness, another well-known quant hedge fund manager, said that while he knew no more about Mr. Simons's methods than anyone else, "It's hard to believe that there isn't a measure of safety in Jim's approach.

"Presumably, he's got a highly diversified portfolio, high turnover, and he's capturing small inefficiencies. It's hard to lose a ton of money doing that. It is always possible that someday his models might stop working. But that's different from 'blowing up.' "

"You know," Mr. Asness added, "human beings have a black box, too. It's called the brain."
As for the new "$100 billion fund," Mr. Simons was even more constrained than usual, thanks to regulatory restrictions that limit what he can say publicly while the fund is raising money. People are buzzing about it nonetheless, for it seems to be a major departure from Medallion. Medallion's investors were almost all wealthy individuals; the new fund, called the Renaissance Institutional Equities Fund, has a $20 million minimum investment and is aimed at institutions. It has a much lower fee structure. It will invest in - or sell short - only publicly traded equities. Instead of making rapid-fire trades, it will be much closer to a buy-and-hold portfolio. And so on.

In one critical way, though, it is similar to Medallion. As the marketing document, which I obtained from a person unconnected to Mr. Simons, put it: "The company's risk control, variance and covariance estimation, execution techniques, slippage models, and predictive signals are all derived from those employed by the managing member in trading the Medallion Funds."

In other words, Mr. Simons believes that computer models similar to those that have worked for Medallion will also work for a fund that can hold $100 billion worth of stocks over long periods of time. It is absolutely audacious.

What interested me most of all was: why? At an age when most men are contemplating retirement, with more money than he can count, why was Mr. Simons still at it? "I enjoy the challenge," he replied.

He then began describing a demonstration he saw recently of a new nuclear accelerator at the Brookhaven National Laboratory, where he is on the board. Two atoms hurtled toward each other, colliding with great force. "A huge number of particles are thrown out," he said, "and the job is to analyze everything that results from the collision."

"Watching the spray of particles on the screen made me think of the stock market," he continued. Every trade, even of a hundred shares of a company, affects every other trade. And every day there are thousands upon thousands of such trades, all of them affecting the rest of the market. His work, as he sees it, is to analyze that incredibly complex mosaic and try to figure out how it all fits together.

"The subject may not be the most important in the world," he concluded, "but the dynamics of the market are really interesting. It's a serious question."

I suddenly understood the motivation behind Mr. Simons's new fund. He's doing it because he wants to see if it can be done. Once a scientist, always a scientist.

Derman: How I became a quant

Emanuel Derman reviews the new book HOW I BECAME A QUANT by Richard R. Lindsey and Barry Schachter.

Nice quote on Derman's blog here:

It always seemed to me, and recent occur[r]ences seem to confirm it, that most algorithmic trading strategies are long volatility but short volatility of volatility.

A previous post from this blog: On the volatility of volatility

WSJ: In 1985, when I left academia and began putting my physics training to work on Wall Street, I talked eagerly about options theory to anyone who would listen. One lunchtime, I turned to a colleague in the elevator and began to babble about "convexity," a mathematical property of options crucial to the Black-Scholes theory used in derivatives pricing. My friend clearly understood convexity, but he shuffled his feet uncomfortably and quickly changed the subject. "Hey, futures dropped more than a handle today!" he said, imitating a genuine bond trader. It didn't take me long to recognize the source of his discomfort: I had just outed him as a fellow quant. Except back then we practitioners of quantitative finance didn't refer to ourselves as quants. That's what "real businesspeople" -- traders, investment bankers, salespeople -- called us, somewhat pejoratively.

Now the term is proudly embraced, as demonstrated by "How I Became a Quant," which collects 25 mini-memoirs of academics who successfully made the jump to Wall Street. Quantitative finance might have lost a little of its luster in recent weeks with the sub-prime mortgage meltdown and its subsequent deleterious consequences for quantitative trading strategies, but quants know -- as many of them in this book emphasize -- that however science- and math-based investment calculations might be, there is still an art to their use and plenty of room for error.

But definitions first. What is a quant, or, rather, quantitative finance? It is an interdisciplinary mix that combines math, statistics, physics-inspired models and computer science, all aimed at the valuation and management of portfolios of financial securities. In practice, for example, a quant might be presented with a convertible bond being issued by a corporation and, by extending the Black-Scholes model to convertible securities, calculate its probable value. Or he might develop a quantitative algorithm to buy theoretically cheap stocks and short theoretically rich ones.

By my reckoning, several of the 25 memoirists in "How I Became a Quant" are not true quants, and they are honest (or proud) enough to admit it. But many others are renowned in the quant community. To name just a few: Ron Kahn, co-author of the classic "Active Portfolio Management"; Peter Carr, an options expert at Bloomberg; Cliff Asness, one of the founders of AQR Capital; and Peter Muller, who ran statistical arbitrage at Morgan Stanley.

Most of the book's contributors belong to the first wave of a financial revolution that began in the 1970s, when interest rates soared, listed equity options grew popular and options traders began to rely on the mathematically sophisticated Black-Scholes model. Investment banks needed mathematical talent, and, as the academic job market dried up, physicists needed jobs. Many early quants were therefore physicists, amateurs who had happily entered a field that didn't yet have a name.

Today we are in the middle of a second wave. As markets became increasingly electronic-based, asset and hedge-fund managers began to embrace algorithmic trading strategies -- and started competing to hire quants, hoping to emulate the continuing successes of such firms founded in the 1980s as Renaissance Technologies and D.E. Shaw & Co. The establishment of the International Association of Financial Engineers, co-founded in 1992 by another contributor to this book, Jack Marshall, has further legitimized the field. Nowadays you can pay $30,000 a year or more to get a master's degree in the subject. Financial engineering has become a profession, and amateurs are sadly passé.

Most of the early quants -- in addition to physicists, they included computer scientists, mathematicians and economists -- came to the field by force of circumstance. Even if they had been fortunate enough to find a secure academic position, they often became weary of the isolating academic grind and found that they liked working at investment banks and financial institutions. As former SAC Capital Management quant Neil Chriss notes, Wall Street is no more competitive than academia. Life in finance is often more collegial than college life itself -- and more stimulating. It is impressive how many of the contributors here cite with awe their encounters with the late economist Fischer Black (1938-95), himself a Ph.D. in applied mathematics rather than economics, who always insisted that research on Wall Street was better than research in universities.

The memoirs in this book are not quite representative. That there are only two women contributors is proportionately accurate; most quants were male. But most quants were also foreign-born. When I ran an equity quant group in the 1990s, the great majority -- all with doctorates -- were from Europe, India or China. Only two of the memoirists grew up abroad in non-English-speaking countries. Quants in the second wave are still largely foreign-born, but more are women and fewer hold doctorates.

Several contributors to "How I Became a Quant" stress an essential point: Physics and finance are only superficially similar. While theoretical physics captures the essence of the material world to an accuracy of 10 significant figures, theoretical finance is at best an untrustworthy, limited representation of the mysterious way in which financial value is determined. Yet Thomas Wilson, the chief insurance risk officer of the ING Group, wisely remarks: "A model is always wrong, but not useless." Despite the inadequacies of quantitative finance, we have nothing better. And, on the practical side, Andrew Sterge, the chief executive of AJ Sterge Investment Strategies, writes: "The greatest research in the world does no good if it cannot be implemented."

Quants do get more respect these days, because their imperfect models can generate profits when used with a knowledge of their limitations. But quants can also produce awe-inspiring disasters when they begin to idolize their man-made models. Nevertheless, most quants, unless they have their own operations, are still second-class citizens on Wall Street rather than its superstars, and many still aspire to leave behind bookish mathematics and join the ranks of the "real businesspeople" who used to look down on them.

How you got here

Below are the top keyword searches that led people to this blog in the past 24 hours. Apparently my post on the time travel movie Primer is popular. Other topics of interest: China, hedge funds, globalization, financial bubbles and the occasional wormhole...

Num Perc. Search Term
10 25.00% primer movie
3 7.50% china climbing
2 5.00% single-digit millionaires
2 5.00% why not hyperinflation
2 5.00% primer/movie
1 2.50% emerging markets debt processing
1 2.50% analyzing hedge fund returns
1 2.50% information processing in brain
1 2.50% advantages and disadvantages of globalization
1 2.50% ltcm today
1 2.50% horvitz cleveland
1 2.50% daniel kahneman and amos tversky
1 2.50% jeremy grantham 2005 interview
1 2.50% john d gartner the hypomanic hedge
1 2.50% cds implied volatility
1 2.50% asness bubble logic
1 2.50% tennessee candidate issues
1 2.50% china-japan relations
1 2.50% all about wormholes
1 2.50% cdx index
1 2.50% china explaining high savings
1 2.50% price to rent ratios
1 2.50% sony vs samsung tv
1 2.50% global real interest rates
1 2.50% why long bonds
1 2.50% price to rent ratio

Blame the quants!

I can already see who the scapegoats will be for the subprime credit meltdown...

The mathematical models involved here are used to value bundles of mortgages or other debt, including corporate "junk" (high yield) bonds. Most importantly, they predict probabilities or rates of default based on historical data and the characteristics of the overall economy, the borrowers, etc. One problem is that rating agencies such as Moody's and S&P were willing to rate senior tranches of subprime debt as AAA (safe), based on the model predictions. In other words, their models predicted that only the riskier tranches would take significant losses and sufficiently senior tranches were as safe as, well, T-bills.

Now, the failure of default models based on historical data might have something to do with loosening of credit standards and outright fraud at the mortgage broker (mainstreet) level. That has little to do with eggheads and math, although perhaps the eggheads should have realized the frailty of human nature in advance :-) Also, there is some question as to whether S&P and Moody's were happy to nudge the ratings higher in order to drum up business. It is an inherent conflict of interest that ratings agencies are paid to generate ratings!

The second model problem is more subtle and plays a role in hedge fund strategies. The models predict relative changes in valuation in different tranches. If interest rates spike, or spreads change, the effect on the senior tranches might be very different from that on junior tranches or on equities. Hedge funds made bets on the correlations predicted by their models, but at least in the short term got into trouble because the market was indiscriminate in marking down all forms of credit, independent of quality. These trades may, in the long run, be big winners if the hedgies have sufficient liquidity to ride them out. Goldman (and some smart co-investors like Hank Greenberg and Eli Broad) and Citadel may have the brains, guts and liquidity to ride this out.

The credit industry is in the early stages of building a system to redistribute risk. This works quite well for us in, e.g., the insurance industry. But it would be naive to think that there won't be hiccups and crises along the way. At the moment, much of the problem is fear and contagion: the system is new and untested, and the participants are afraid.

Final comment: I doubt the typical market neutral quant long-short fund is directly involved with credit products. They lost money recently simply because the market moved in a very unpredictable way -- certain funds that did have credit exposure had to sell whatever liquid positions they had to make margin calls. That means stocks that quant models tended to favor suddenly and unexpectedly went way down...

For Wall Street's Math Brains, Miscalculations

Complex Formulas Used by 'Quant' Funds Didn't Add Up in Market Downturn

By Frank Ahrens
Washington Post
Tuesday, August 21, 2007; A01

They are the powerful, cerebral and offstage actors of Wall Street, but the recent turmoil in the financial markets has yanked them into the light.

They are the math geniuses of the quant funds.

Short for "quantitative equity," a quant fund is a hedge fund that relies on complex and sophisticated mathematical algorithms to search for anomalies and non-obvious patterns in the markets. These glitches, often too small for the human eye, can present opportunities for short-and long-term trades that yield high-profit returns.

The models replace instinct. They try to turn historical trends into predictive science, using elegant mathematics seemingly above the comprehension of your average 401(k) participant or Wall Street fund manager.

Instead of veteran, market-savvy traders waving fistfuls of sell slips, the elite quant funds employ Nobel nerds with math PhDs, often divorced from the real world. It's not for nothing that they are called "black-box" funds -- opaque to outsiders, the boxes contain investment magic understood by only the wizards who conjured it up.

But the 387-point drop in the Dow Jones industrial average Aug. 9 and the continuing turmoil in the markets, in part attributed to massive sell-offs by the quant funds, have tarnished some of the quants' glimmering intellectual credentials and shown that, when push comes to shove, they can rush toward the exits as fast as a novice investor.

Last week, Goldman Sachs said its Global Alpha quant fund had lost 27 percent of its value this year because its computers failed to anticipate what the firm called "25 percent standard deviation moves" or events so rare Goldman had seen them only twice before in the firm's history. On the same day Goldman revealed the bad news, the firm said it would lead a group of big-money investors, including philanthropist Eli Broad, in pouring $3.6 billion into another Goldman quant fund, aiming to shore up confidence in the quants.

Barclays Global Investors, with $450 billion of its $2 trillion in assets under quant management, began applying mathematical tools to its funds in 1978. Last week, Barclays spokesman Lance Berg said the firm was "maintaining its investment process" despite the recent troubles. He would not say how much the Barclays quant funds had fluctuated during the period of turmoil.

The acknowledged quant king is James Simons, 69, an M.I.T.-trained mathematician with a groundbreaking theory that physicists are using to plumb the mysteries of superstring study and get at the very nature of
existence itself. Simons turned his big brain on investing after his math career, founding Renaissance Technologies quant shop. The firm pocketed $1.7 billion in investor fees last year, among the highest in the industry. In return, his clients can reap annual returns of more than 30 percent, according to news reports.

As elegant as the models are, they cannot predict unpredictable events, or human panic, some traders say. Further, some say, too many quant funds are full of myopic brainiacs, overly reliant on their tools.

"Most are idiot savants brought to industrial proportion," Nassim Nicholas Taleb, former quant-jock and bestselling contrarian author, said by phone from Scotland, where he is promoting his new book on improbability, "The Black Swan."

"They are very smart in front of a textbook but not smart enough to understand very elementary things in reality," he said.

Taleb believes in monkey-wrench events that shatter the models of the quant-jocks. He says their algorithms don't adequately account for huge, rare anomalies, such as the current surprise credit crunch. Or the Russian credit crisis in 1998 that nearly put the superstar quant fund of the time, Long-Term Capital Management, out of business in a matter of days, saved by cash infusion organized by the Federal Reserve.

The sentiment is reminiscent of the demise of Enron, a company said to have been designed by geniuses but run by idiots. The oil-and-gas trader used next-generation financial tools designed by brilliant mathematicians. But they couldn't overcome the inept and criminal actions of the management.

The allure of a unifying, perfect mathematical formula is powerful; it is an alchemy for the enlightened age. Math's universal principles underlie and suffuse everyday life and the workings of the cosmos, offering a glimpse of the eternal. In the frequently irrational financial markets, mathematic models offer the hope of cool reason and certitude, a sort of godlike wisdom.

In the 1998 film "Pi," a troubled math genius who sees patterns in the newspaper stock tables tries to create the Algorithm for Everything. He and his work are simultaneously hunted by a Wall Street firm that seeks its predictive powers, and by orthodox Jews, who believe it could unlock the mind of God.

The quant funds thrive on volatility -- it's how they make their profit margins. But recent weeks have proved too volatile for some of the funds, many of them highly leveraged, which seemingly all at once got spooked into seeking liquidity. When they ended up seeking liquidity by selling the same stocks, the Aug. 9 plunge happened, analysts speculate, resulting in the Dow's second-largest one-day slump of the year.

"It became increasingly transparent that many of the highly sophisticated quant funds employed similar investment approaches and held similar core holdings," Thomson Financial wrote in an analysis of the role of the 25 largest quant funds in the market meltdown. "This resulted in the funds selling similar long stocks and covering similar short positions."

For instance, the most broadly held stock among the top quant shops, Thomson reported, is Exxon Mobil. Shares of the oil company dropped 2.4 percent in heavy trading during the Aug. 9 sell-off.

"If you ask the question, 'Did the smart guys blow it or get it right?' I think the answer is, if they knew it, it wouldn't have happened," said David Levine, a vice president in corporate advisory services at Thomson.

"I occasionally hear broad statements like, 'This just shows computer models don't always work,' " Clifford S. Asness, founding principal of the quant-fund firm AQR Capital Management, wrote to his clients after the sell-off. "That's true, of course, they don't, nothing always works. However, this isn't about models, this is about a strategy getting too crowded, as other successful strategies both quantitative and non-quantitative have gotten many times in the past, and then suffering when too many try to get out the same door."

The value of Simons's $29 billion Renaissance Institutional Equities Fund fell by nearly 9 percent from the beginning of the month through the Aug. 9 drop, Bloomberg News reported. It was less of a hit than many of the other quants took, possibly reinforcing Simons's status as the Dumbledore of the quants.

A mathematician and cryptanalyst, Simons headed the math department of the State University of New York at Stony Brook, pushing the program into the nation's elite.

Simons and his colleagues work in a form of high math decipherable to a handful of humans on the planet. As such, practitioners of the rare mathematic arts can become the powerful priests of investing, thanks to their strange and obscure language, much the way the medieval church trafficked in Latin, which required the translation of a learned cleric.

In 1978, Simons began to apply his predictive models to investing and set up his investment shop on the north shore of Long Island near his old school, virtually insulated from Manhattan's financial district. He generally recruits mathematicians and programmers, not MBAs and traders.

The press-shy Simons would not comment for this article, and a Renaissance spokesman could not be reached for a comment.