Algo trading raises stakes for market regulators
Mon Jun 4, 2007 10:01AM EDT
By Kevin Plumberg
NEW YORK, June 4 (Reuters) - As investors increasingly use hyper-fast computer programs to trade, financial market regulators have begun using similar software, fearing that if misused the technology could have dire consequences.
The use of algorithms, or algos, to make complex decisions and place thousands of orders in milliseconds has grown in popularity, particularly among equity and currency dealers.
But the combination of heavy order volume from high frequency investors and the rapid pace of trading has created fertile ground for mercenary-like dealers who use technology and aggressive tactics to steamroll rivals.
Market regulators are eyeing the potential risks closely.
"The greatest difficulty with tracking these systems is the risk management aspect," said Edward Dasso, director of trade practice and market surveillance at the National Futures Association, a self-regulatory organization for the U.S. futures industry.
"They can pump in more orders per millisecond than a person can, so there's always the concern that they can melt a firm and blow it up in a matter of seconds," Dasso said.
Think of a rogue trader like Nick Leeson, whose $1.4 billion of losses caused the collapse of Barings Bank in 1995. Then imagine a character like him with automated programs that could send thousands of orders per second.
"Whenever you have computers generating orders and trading versus a human, that is something we as regulators have to worry about," said Tom Gira, executive vice president with the National Association of Securities Dealers.
One of the biggest concerns about a Wall Street in which machines increasingly are making million-dollar decisions is that the market has not yet been significantly stress-tested.
In the current era of very low volatility, a single bad trade could potentially set off a chain reaction of bad trades as herds of investors rush to exit a sinking market. And because of the feverish pace of algo trading, the losses could whip through multiple markets in a split second.
A similar scenario of systemic risk happened over a longer period of time in 1998 when the hedge fund Long-Term Capital Management nearly collapsed because of several trades gone bad, and threatened the entire U.S. financial system.
The Federal Reserve was forced to engineer a $3.6 billion bail-out of LTCM to prevent bigger damage to investment banks that supplied the fund with leverage.
Gira from the NASD said it's possible that the proliferation so-called "black-box trading" could actually help prevent such a system-wide breakdown since the initial bad trade would be almost immediately acted on as a trading opportunity -- and because of the sophistication of tools available to regulators now.
The NFA uses what Dasso described as automated systems for surveillance purposes, and the NASD utilizes pattern recognition tools. A major U.S. financial regulator also has closed a deal to use an algorithm platform from Progress Software, according to sources.
However, the nature of algorithmic trading makes sleuthing tricky for market regulators because high frequency investment firms will often run several algos at once with strategies that may have conflicting ends.
So it is possible that a firm could be on both sides of a trade simultaneously without even knowing it. "The key is understanding the different strategies in place," the NFA's Dasso said.
COLD WAR
Longer-term institutional investors tend to use algorithms to disguise their trading of large blocks of shares, test strategies and identify arbitrage opportunities, while short-term investors are more focused on rapid-fire trading.
Many users of algorithms are notoriously secretive about their strategies and will often change them daily or even within a session to stay competitive.
In fact, industry participants say that investors and traders using algos will sometimes run programs to track what other algos are doing.
For example, if an algo is buying a large block of a company's shares by slicing the block into little pieces and buying in intervals, a rival's black box may be able to front-run the orders by discerning the pattern and purchasing right before the other algo.
The likely result: The rival buys the security at a lower price and can make a profit as the activity drives the price up, while the initiator of the trade loses out by paying higher prices.
"It's like the Cold War," said John Bates, founder of Progress Apama event processing technology. "People are sending out submarines and sometimes they'll see other submarines and follow them and track them," he said.
Of specific concern to the NASD's Gira is if a broker dealer tracks the way clients are using a platform offered by the firm, in order to trade off the information.
Front-running by traders, which is illegal if based on information about customers' orders, has been a major headache for regulators for many years but the increasing power of the machines could make it more profitable far faster and more difficult to discern.
"Information leakage is an area we are concerned about, particularly when it comes to algorithmic trading," Gira said.
DARK POOLS
Another challenge for regulators watching algos has to do with an area off the exchanges. For example, at some banks there are trading platforms that anonymously match buy and sell orders for financial assets, known in market parlance as "dark pools" of liquidity.
The attraction of dark pools is mainly to automated systems that don't require a price quote and want to keep costs of trading down. However, the dark pools lack the transparency of an exchange.
Tabb Group expects dark pools will increase their daily volume to 1.5 billion shares in 2010 from 512 million shares this year. But the risks will remain profound.
"In the dark pools, it becomes easier for wolves to wear sheep's clothing," said Tabb Group analysts Jeromee Johnson and Larry Tabb in a recent report.
For example, proprietary traders could prowl the pools and do small trades to detect institutional investors selling or buying blocks of shares. The traders would then use the information gained through these trades against others who didn't have it -- a practice known as "toxic liquidity."
Sang Lee, research director with the consulting firm Aite Group, said that more than ever it will be up to brokers to police their own pools. Their biggest incentive is to avoid being made an example by regulators if widespread improper trading is found in one.
But how effective that will be or how willing investors will be to keep the pools free of improprieties is in doubt.
"People are going to do the bare minimum to meet the regulations or what they think the regulations are," said Jeffrey Hudson, chief executive of technology vendor Vhayu, at a recent industry meeting. |
