Vanishing Crystal Ball Gives Way to the Best Black Box in the Industry

January 9, 2001

Book Review
James Case

The Predictors: How a Band of Maverick Physicists Used Chaos Theory to Trade Their Way to a Fortune on Wall Street. By Thomas A. Bass
Henry Holt, New York, 1999, 310 pages, $25.00.

Thomas Bass, who has written for The New Yorker, Wired, and other top-shelf periodicals, knows a good story when he sees one. His first book, The Eudaemonic Pie, chronicled the famous assault of Doyne Farmer, Norman Packard, and other charter members of the "Chaos Cabal" at UC Santa Cruz, using toe-operated computers concealed in their shoes, on the roulette tables of Las Vegas. The book under review amounts to an informal corporate history of the Prediction Company, founded by the same dynamic duo, with the stated intention of making enough money to finance (their own and other) future research. Bass was present at the firm's foundation and remained in touch throughout its first seven years. The result toggles between "up close and personal" sketches of those involved and brief nontechnical descriptions of their scientific deeds.

The company grew out of a series of three interdisciplinary conferences on economic systems held at or near the Santa Fe Institute between August 1986 and February 1991. Each was intended to bring a slightly different group of economists and financiers together with physical and computer scientists in the hope that quantitative methods developed by the latter for the study of "complex adaptive systems" might prove useful to the former. The first of the three conferences was small, took place at a neighboring dude ranch, and lasted only two days. It was followed by a larger one, held at the institute itself, and chaired by two Nobel laureates. Philip Anderson (an authority on condensed-matter physics) and Kenneth Arrow (a theoretical economist) brought together a dozen or so economists with a roughly equal number of physical and computer scientists for two full weeks.

Interactions between the two groups were often contentious, in large part because the economists and financiers were unable to persuade the physical and computer scientists that their assumptions were justified. Experienced users of mathematics, the physicists and computer scientists were quick to spot the dubious nature of the axioms from which mainstream economic theory deduces its celebrated conclusions. A consensus soon developed on their side of the table that even the most mathematical of economic theories assume the greater part of what they ought to prove. And the more they badgered the economists during the question-and-answer periods, the more defensive the latter became. The conference proceedings, published the following year, described the meeting as "a serious dialogue between disparate, and at times hostile, communities of scholars."

The third and last conference in the series, Wall Street and Economic Theory: Prediction and Pattern Recognition, was held two and a half years later. Its comparatively practical focus is reflected by the Wall Street traders and investment bankers who, for the most part, had replaced the academic economists on the market-oriented side of the table. Surely, the physical and computer scientists thought, these fiscally responsible agents---who daily expose billions of other people's dollars to what they deem prudent risk---would be able to offer a more plausible account than their academic colleagues of the inner workings of the global economy. But to the considerable surprise of the physical and computer scientists, most of the market professionals also professed to believe some version of the classroom teaching known as the "efficient-markets hypothesis," whereby the latest price quotations are held to represent the best possible estimates---given current information-of the values of listed assets.

Although investors as prominent as Warren Buffet and George Soros have denied---repeatedly and emphatically---the efficiency of financial markets, an assertive minority of those present kept bringing the discussion back to the efficient-markets hypothesis, and to the alleged impossibility of predicting efficient markets. Finally, unable to contain himself any longer, Farmer pointed out that during the first day of the conference, a number of market professionals had detailed their systematic exploitation of market inefficiencies. So why not stop debating whether it was possible to do what they had so obviously done and start figuring out how others might do it too?

The meeting did nothing to weaken the conviction Farmer shared with Packard---and several others present---that dynamical systems research can predict at least some of the movements of the financial markets. On the contrary, it strengthened their resolve to demonstrate the power of such methods!

The Prediction Company was officially founded on September 12, 1991, in Santa Fe, New Mexico, by three senior and five junior members. The senior members were Farmer, Packard, and James McGill, a former guitar player, surfer, and physics grad student at UC Santa Cruz who, shortly after graduating, had abandoned scientific pursuits in favor of venture capitalism. Also present was James Pelkey, a long-time business associate of McGill's, who had recently moved to Santa Fe. Pelkey's once-promising financial career, launched with a Harvard MBA-to complement his engineering degree-had skidded off track when his former wife shot him. The bullet in his spine left him in constant pain and paralyzed from the waist down, but failed to diminish his enthusiasm for high-tech start-ups. It would be difficult to overstate the importance of his decision to back the Prediction Company for several months, until longer-term funding could be arranged.

After much deliberation, it was decided that each senior member of the firm should own three times as many shares of stock as each junior member, on the ground that---although eventual success or failure might depend on the quality of the work done by the junior members---the initial ability to raise money would depend exclusively on the reputations of the senior members and the credibility of the business plan they would be presenting to potential investors. Not without difficulty did Pelkey and McGill convince the junior members that a more egalitarian distribution would scare off potential investors: Experienced "money men" are known to shun situations in which the fortunes of the "key players" are less than securely entwined with those of the firm.

A running joke throughout the book concerns the "company suit," which was purchased with company funds and stored in the corporate offices because neither Farmer nor Packard owned attire suitable for calling on venture capitalists. First one and then the other wore the suit to perform the dog-and-pony show they developed for the benefit of potential investors to whom McGill introduced them. Bass sounds almost like a Hollywood gossip columnist when describing the shirts, shoes, and neckties (Farmer favored a tie adorned with M.C. Escher's famous "flying goose" pattern) with which they completed their fund-raising ensemble. At various times, they managed to elicit the interest of Merrill Lynch, Paine Webber, Salomon Brothers, Montgomery Securities, Bank of America, and Kidder Peabody.

Even the fact that the firm was firmly rooted in remote Santa Fe---several hours' flying time from the nearest financial market---did not prove fatal. The wonders of electronic communication were already well known in the investment community, and the Prediction Company had been quick to demonstrate its ability to acquire information when and as needed. Moreover, Bass makes a point of describing firm members' devotion to the Rocky Mountain lifestyle: Every meeting seems to have been convened on a Friday, and to have adjourned at a civilized hour to a nearby pub. All in attendance seem to have had plans to spend the coming weekend hiking, biking, or kayaking some nearby natural wonder.

Among the first potential backers approached was Edward Thorpe, inventor of the first successful card-counting strategies in blackjack. While employed as a junior (mathematics) instructor at MIT, Thorpe used the school's mainframe computer to conclude that a player endowed with perfect recall, and using a strategy devised by Thorpe, would enjoy an advantage of roughly 3% over the house. Beat the Dealer, his subsequent best seller, chronicled an actual field test of his system, conducted in Las Vegas under the watchful eye of a reporter from Life magazine. Thorpe calculated at the time that he could earn about $300,000 a year by playing the game full time, but soon realized that financial markets constituted a far more promising field of application for his particular skills.

Thorpe later explained that "Wall Street is like a big gambling casino" in which "the game is much bigger and much more interesting to me than casino gambling." But he began with only the faintest idea of how to go about winning on Wall Street. The requisite suggestion came from Sheen Kassouf, an economist he met soon after moving to UC Irvine. His new acquaintance pointed out that certain investments---a.k.a. bets on the performance of particular stocks---could be hedged via the purchase of "warrants" (essentially options) against those same stocks in such a way that the purchaser would come out ahead whether the relevant stock prices went up or down.

In those days, before the publication of the Black-Scholes option-pricing formula, warrants were often grossly mispriced. Thorpe soon developed a computer program to identify such opportunities; its deployment was so successful that, by 1970, both Thorpe and Kassouf had abandoned academe for greener pastures. Thorpe listened with interest to Farmer's pitch but declined to become finan-cially involved. Having by then spent upward of twenty years exploiting the inefficiencies of financial markets, he no longer felt the need to prove that it could be done.

The Prediction Company found the "angel" it needed in the person of David Weinberger, described by Bass as "the original Wall Street rocket scientist." Having written a thesis on combinatorial transversality under Ray Fulkerson at Cornell during the early 1970s, Weinberger spent a few years working at Bell Labs and then teaching computer science at Yale, before seeking employment on Wall Street. He was fortunate enough to attract the interest of a youthful partner at Goldman Sachs. As head of the options area, Robert Rubin (who would later become treasury secretary in the Clinton administration) was able to offer him an entry-level position---contingent on the outcome of psychological testing-trading junk bonds. In 1976, even Goldman Sachs doubted the sanity of anyone willing to abandon academic life for Wall Street.

Weinberger soon exhibited a golden touch as a bond trader. He was among the first to flourish in the market for derivative securities, and he is credited-at least by Bass-with the invention of stock index arbitrage. (The latter consists of buying or selling futures in (say) the Dow Jones Industrial Average in one market, while buying or selling the compo-nent shares in another.) When Wein-berger was hired away by the Chicago firm of O'Connor & Associates, his replace-ment in the equity department at Goldman Sachs was Fisher Black, co-inventor of the Black-Scholes option-pricing model.

By 1986, Weinberger had become co-managing partner of O'Connor, with effective command of the entire operation. Within five years, however, the firm was to be bought out by the Swiss Banking Corporation (SBC). Since the buyout would disband Weinberger's pet project---the in-house research department---part ownership of a remote research facility became attractive to him. As a result, O'Connor & Associates signed a five-year development contract with the Prediction Company on September 29, 1992, just a day before being itself acquired by SBC.

At this point, barely a year after incorporation, the predictors were still engaged in "shadow trading." This consisted of recording for later evaluation the deals the firm's proprietary models would have directed it to make, had there been money with which to make them. Everything appeared to be working brilliantly. But as soon as the flow of "live" recommendations began, in accordance with the new SBC contract, the company's crystal ball seemed to vanish. Instead of making money on paper, they began losing it in reality.

During the recriminations that ensued---in which several junior members of the firm expressed dissatisfaction with the leadership provided by the senior members---it became clear that ange was inevitable. After much soul searching, the senior members announced that two of the five junior ones would have to leave. Each received two months' severance pay, and some assistance in finding a new position, along with his accumulated shares of stock.

By the end of 1995, McGill had also decided to leave the firm, along with the last of the founding junior partners. Yet business was picking up. SBC extended the development contract through the end of 1999, and experts in programming and (market) data acquisition were found to replace the converted physicists who had left. The latter had tended to regard coding and data management as chores, and to delay doing them. Since the models they built were data-hungry in the extreme, they managed to launch the company on a collision course with their own limitations. The result was inevitable. Like so many other activities, the polyglot they jokingly referred to as "phynance"---the marriage of physics and finance---had begun to demand specialized skills.

The book follows the firm through the end of 1998. By then the payroll had expanded to include 20 full-time employees, forcing a move to larger quarters. The Prediction Company now occupies the former Capitol Hotel in downtown Santa Fe, complete with 14-foot ceilings on the first floor, a second-floor kitchen, admirable views of the surrounding mountains, and an outdoor deck overlooking a liquor store and a jazz club. The building has been rewired from pillar to post and houses what some describe as "the best black-box in the industry," meaning that the firm's ancillary management tools---for data acquisition, risk analysis, portfolio review, and the like---are among the best in the business. Other firms seem anxious to acquire them. As the book went to press, Packard had taken over as CEO, leaving Farmer free to focus on models and model development. Save for becoming filthy rich, and without abandoning a life style they obviously cherish, the two seem to have accomplished the greater part of what they intended.

James Case writes from Baltimore, Maryland.

Renew SIAM · Contact Us · Site Map · Join SIAM · My Account
Facebook Twitter Youtube