Wednesday, March 10, 2010

Patterson, The Quants

It’s all too easy to nitpick over Scott Patterson’s The Quants: How a New Breed of Math Whizzes Conquered Wall Street and Nearly Destroyed It (Crown Business, 2010). But why bother? It’s a thoroughly enjoyable book, and how many of those have you read lately? We get to know “the players,” as Patterson dubs them, warts and all: Peter Muller, Ken Griffin, Cliff Asness, Boaz Weinstein, Jim Simons, Ed Thorp, Aaron Brown, Paul Wilmott, and Benoit Mandelbrot. Some had skin in the game, others were Cassandras. We’re also introduced to some of their strategies: delta hedging, stat arb, value/momentum, high frequency trading, and now and then sheer moxie.

The book is a cautionary tale. Grossly simplified, it argues that although quantitative strategies have the deceptive allure of science, financial markets are unlike the physical world and cannot be accurately modeled mathematically. Moreover, quantitative strategies are easily replicable, at least in outline, leading to overcrowded trades. When volatility spikes and liquidity dries up, every quantitative strategy will get crushed. In August 2007 even Renaissance’s Medallion fund took a drubbing.

Quants are not going away any time soon, nor should they. But they (and, more important, their risk managers and, most important, those who manage the risk managers) need to recognize the limitations of their models and strategies. It’s not just those fat tails that wag the financial markets. One of the most obvious problems, stemming in part from the very nature of quantitative analysis, is the copycat phenomenon. That is, it’s easier to copy or reverse engineer a math formula or a piece of computer code than a judgment about the quality of a firm’s management. Just consider the recent rush to reap the rewards of high frequency trading. The problem is that as more and more players enter the same business profit margins tend to get squeezed. And then entrepreneurs, not necessarily quants, are inclined to ramp up risk, whether by increasing leverage or some other strategy, to keep the profits flowing. They try to hit home runs instead of singles and doubles, generally a recipe for failure.

What keeps good quant funds going, continuing to make money even in difficult times? Intelligence, originality, hard work, and luck would probably head the list. Paranoia might also help. For instance, Jim Simons and his merry band of math and science Ph.D.s allegedly devoted the first forty hours of their work week to assigned tasks; during the “second forty hours” they could experiment, with full access to all of the fund’s data. Their intelligence, originality, hard work and, as Simons has always said, luck have kept the Medallion money machine working in overdrive throughout a range of conditions (except in August 2007 when they were the only buyers in town, a very lonely and expensive scenario). In terms of strategy, what sets Medallion apart? We learn only (but here “only” might be worth a fortune) that it adjusts for changing market conditions far more frequently than its competitors.

In general Scott Patterson’s book gives up few trade secrets, but that’s just as well. Riding the coattails of the super successful might be profitable much of the time, but who wants to be there for the multi-trader pileup? What The Quants offers the wannabe alpha trader is a series of portraits of very bright, passionate men, most with oversized egos, who are driven to win. We see them both on the way up and when they’re teetering on the brink. We may not like them all, but I for one was transfixed reading about them.


  1. Interesting review.

  2. I just finished reading the book this weekend and learned a lot of newer terms...The Truth, "mathletes", and enjoyed it almost as much as Bailout Nation and Too Big To Fail.
    Nice site Dr., I put that holistic reading list of yours into my favorites.
    Interesting byline in the book about how a man was sent to a computer room and left perplexed a couple times as he only saw a woman and assumed she could not be a quant.