|
Thanks to slumping markets, investment banks are shedding many of their highly-paid traders. When markets recover, the banks might be tempted to replace them with rather cheaper talent. One alternative has been around for a while but has yet to catch on: autonomous trading agents-computers programmed to act like the human version without such pesky costs as holidays, lunch breaks or bonuses. Program trading has, of course, been done before; some blamed the 1987 stockmarket crash on computers instructed with simple decision-making rules. But robots can be smarter than that. Dave Cliff, a researcher at Hewlett-Packard Laboratories in Bristol, England, has been creating trading robots for seven years. In computer simulations he lets them evolve genetically , and so allows them to adapt and fit models of real-world financial markets. His experiments have suggested that a redesign of some markets could lead to greater efficiency. Last year, a research group at IBM showed that Mr Cliffs artificial traders could consistently beat the human variety, in various kinds of market. Nearly all take the shape of an auction. One well-known type is the English auction, familiar to patrons of the salesrooms of Christies and Sothebys, where sellers keep mum on their offer price, and buyers increase their bids by stages until only one remains. At the other extreme is the Dutch auction, familiar to 17th-century tulip-traders in the Netherlands as well as to bidders for American Treasury bonds. Here, buyers remain silent, and a seller reduces his price until it is accepted. Most markets for shares, commodities, foreign exchange and derivatives are a hybrid of these two types: buyers and sellers can announce their bid or offer prices at any time, and deals are constantly being closed, a so-called continuous double auction . | 
