NEW YORK (AP) - If you saw a penny on the sidewalk, would you pick it up?
You may think it's not worth the effort, but a breed of investors who have been in the news do. Using super-fast computers, high-frequency traders in effect bend down to pick up pennies lying about in the stock market - then do it again, sometimes thousands of times a second.
More than a week after the Dow Jones industrial average fell nearly 1,000 points, its biggest intraday drop ever, regulators are still sifting through buy and sell orders to figure out what sparked it. One big focus are orders placed by high-frequency traders, or HFTs, and for good reason. These quick-buck firms barely existed a few years ago but now account for two-thirds of all U.S. stock trading.
In other words, all those TV pictures of the stately New York Stock Exchange building on the evening news are an illusion. The real action on Wall Street is far away in Kansas City, Mo., and in New Jersey, in towns like Carteret and Red Bank, where HFTs named Tradebot and Wolverine and Tradeworx ply their trade.
High-frequency trading firms, which number over 100, use computers programmed with complex mathematical formulas to comb markets for securities priced too high or too low because traders haven't had to time to react to the latest data. The computers then buy or sell in a split second, locking in a profit.
The opportunities seem hardly worth noting. They're not just fleeting, but small, often a penny or less.
But those pennies can add up to a lot of money, enough to draw the attention of Goldman Sachs Group Inc., the giant Chicago hedge fund Citadel Investment and other big financial firms. In recent years they've paid hundreds of millions of dollars for stakes in high-frequency trading companies.
The money has stoked what was already fierce competition among the firms for a leg up.
To spot opportunities and act on them before others, HFTs are constantly hunting for faster computers. They also locate themselves close to the big exchanges' data centers. That can cut their trade times by milliseconds.
One way these traders make money is by exploiting the fact that stock indexes sometimes don't immediately reflect falling or rising prices of their component stocks, said Manoj Narang, chief executive at Tradeworx of Red Bank, N.J. If Microsoft shares rise 5 percent but an index fund that includes it such as the SPDR S&P 500 lags by a fraction of second to adjust, his computers will automatically buy shares of SPDR S&P 500 at the lower price and then sell them again when they are fully valued.
Or maybe Microsoft is trading in London at a penny less than it's trading at the same moment in New York. A high-frequency trader will buy shares in London and wait for them to rise.
Since the discrepancy lasts a mere fraction of a second, speed is key.
Narang boasts it takes only 15 millionth of a second for his computers to place a buy or sell order after detecting an opportunity.
Or, as he puts it, "If you try to pick up the penny, we'll probably beat you to it."
So is that good or bad for the market?
If you listen to HFTs, all their fast trading benefits big and small investors alike. More trading means more bids and asks for shares, and that cuts the time needed to find someone willing to buy what you're selling or vice versa. Costs also fall. With more bids and asks, the difference between the price you seek and the price offered (what traders call the "spread") will likely narrow. You get to keep more of your money.
High-frequency traders see themselves as part of a long tradition of using technology to shake up Wall Street.
For decades an order to buy or sell a security went to a person in a trader's jacket standing on the floor of an exchange, often at the NYSE in Lower Manhattan. If you wanted to sell stock in General Electric, for instance, these so-called specialists would find a buyer. If they couldn't find one, they bought it themselves.
In exchange for their services, the specialists pocketed some of the difference between the price at which you were willing to buy and the price at which a GE holder was willing to sell.
This system came under attack in the early 1980s from Nasdaq, a rival marketplace for stocks, which began using computers to make trades. The pitch was it could match buyers and sellers faster than humans, and for less money.
Then, starting in the late '90s, the NYSE specialists got hit again, this time with a series of blows: new rules encouraging computer matching of buyers and sellers, a shift to quote stock prices in minute increments of decimals instead of fractions, and a decision to cut the minimum spread that specialists or other middlemen could grab for themselves from 6.25 cents per share to a penny.
"It used to be an oligopoly, an old boy's club," said Irene Aldridge, head of an HFT shop called Able Alpha Trading and author of "High-Frequency Trading." "But now it's a completely level field."
Critics of high-frequency trading say all this talk about narrowing spreads for ordinary investors distracts from a key problem: Split-second trading without human supervision is a recipe for disaster
Exhibit A: the May 6 crash.
One theory about the drop is that, unlike the NYSE, the new exchanges and trading networks catering to HFTs didn't apply any "circuit breakers." These are designed to halt trading momentarily during a freefall to stop selling from feeding on itself.
In others words, without circuit breakers the computers went crazy.
Another theory holds that it wasn't quick-fire trading by HFTs that made things worse but a lack of it. Some reportedly pulled back when stocks started dropping, removing liquidity when it was needed the most.
Whatever the answer, this much is true: These secretive firms are likely to grab the spotlight for a while now. And their trading might get even more frenetic.
After the May 6 freefall, all manner of trading rules are up for debate. But it's worth noting that until recently regulators were considering cutting the minimum spread again, possibly to half a penny.
"People will be needing even better computers," said author Aldridge.