Mallaby, More Money Than God

Over the weekend I read Sebastian Mallaby’s More Money Than God: Hedge Funds and the Making of a New Elite (Penguin Press, 2010). I admit that I’m addicted to books on top hedge fund managers; I’m fascinated by how these gurus think—and act. Although Mallaby’s book often retraces some well-trodden ground, the columnist and former editorial board member at the Washington Post writes so compellingly that I was never tempted to skim.

Since this book came out in June and has been reviewed extensively, in today’s post I’ll focus exclusively on the hedge fund manager and team I will never be able to emulate sitting at my Dell Precision workstation but whose consistently high returns remain a goal—James Simons and the mathematicians and scientists who work at Renaissance Technologies. Between the end of 1989 and 2006 Medallion, its flagship fund, returned an average of 39% per year.

One of the reasons the Medallion fund has been so successful is that Renaissance draws on insights from fields other than economics and finance. For instance, the early team was experienced in cryptography and, more generally, in military communication. Elwyn Berlekamp, a Berkeley mathematician and along with Simons and James Ax a veteran of the Institute for Defense Analyses, “had worked on systems that send signals resembling ‘ghosts’—faint traces of code in seas of statistical noise, not unlike the faint patterns that hide in broadly random and efficient markets. Soldiers on a battlefield need to send messages to air cover that are so wispy and translucent that they won’t betray their positions: Not only must the enemy not decode the messages; it must not even suspect that someone is transmitting. To Berlekamp, the battlefield adversaries fooled by such systems bore a striking resemblance to economists who declared markets’ movements to be random. They had stared at the ghosts. They had seen and suspected nothing.” (pp. 286-87)

It took about a decade for Renaissance to begin capitalizing on ghost hunting. Henry Laufer, a mathematician at Stony Brook, “spotted patterns in the way that markets move right after an event perturbs them. … It was not that a commodity would jigger in the same way following every piece of news: That would have been too obvious. But if you scrutinized thousands of reactions to thousands of events, certain sequences emerged in slightly more than half of all the observations.” (pp. 287-88) Renaissance began to replace its underperforming trend-following models with short-term statistically driven trading, and the profits became robust.

“Many of the patterns that Renaissance discovered,” Mallaby writes, “were individually modest…. But by discovering a large number of minor inefficiencies and blending them into a single trading program, Renaissance built a system that racked up profits year after year, especially during periods of turbulence.” (p. 290)

If the early team at Renaissance came from a cryptography background, two later additions (and now co-heads of Renaissance after Simons’ retirement) came from the field of computerized translation. Through their work at IBM they transformed this field, earlier “dominated by programmers who actually spoke some foreign languages. The approach was to understand the language from the inside, to know its grammar and its syntax,” and to teach the computer a language “much as you might teach a middle schooler.” (pp. 298-99) But Peter Brown and Robert Mercer had a radically different method. “… they got hold of Canada’s parliamentary records, which contain thousands of pages of paired passages in French and English. Then they fed the material into an IBM workstation and told it to figure out the correlations.” The computer began to search for patterns, sometimes combing the data multiple times “using slightly different algorithms at each turn.” The results “outclassed competing translation systems by a wide margin.” (pp. 299-300)

Brown and Mercer believed that “the crude force of a computer’s memory can actually substitute for human notions of intelligence and science. And computers are likely to work best when they don’t attempt to reach results in the way that humans would do.” (p. 301) So, contrary to other hedge fund programmers who “trained a machine to approach markets in a manner that made sense for human traders,” the Renaissance duo “trained themselves to approach problems in a manner that made sense for a computer.” (p. 301) Robert Mercer, I suspect proudly, admits that “Some signals that make no intuitive sense do indeed work. … The signals that we have been trading without interruption for fifteen years make no sense. Otherwise someone else would have found them.” (p. 302)