Wednesday, August 22, 2007

For Wall Street's Math Brains, Miscalculations -

For Wall Street's Math Brains, Miscalculations -

"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."

Comment: Mathematical models are not magic. These harness the computer to calculate more and faster than would be possible with the unaided eye and mind. But they are only as good as the model that they embody, the data on which their parameters are estimated, and the quality of the implementation of the theory in the model. One major weakness is that knowledge decays, and a model that was once good at describing the world can lose its power as the world changes. A model that was good at predicting the stock market when volatility was low may not be so good when volatility is high. Even more, a model that is predicated on the assumption that other people will not be doing what it is recommending that its users do, may be outmoded when lots of traders start using very similar models. And so it goes.... JAD

1 comment:

John Daly said...

I used to think that quantified models clarified the analytic process, and they do in the sense that the logic is explicit. Certainly decisions made by individuals depend on lots of implicit knowledge (and misinformation) and can seldom if ever by fully understood. When one gets to decisions made by groups of people, of course, understanding is even harder to come by.

But when you get to complicated mathematical models implemented by computers, they can become very opaque. Even experts may be unable to really understand the key assumptions made in the models, and thus unable to monitor if and when those assumptions may become questionable due to changing circumstances.