This content is from: Home

Hooked on speed

Statistical arbitrageurs and other high-frequency traders demand lightning-quick data access and trade execution that, thanks to technological and regulatory changes, the market can now provide.

DON'T TALK ABOUT GETTING TRADES EXECUTED QUICKLY to Alistair Brown, the no-nonsense Scotsman who runs Lime Brokerage from his TriBeCa office in downtown Manhattan. Speed is Brown's calling card. He has built his 35-person firm with computer programmers, finance gurus and the cream of academia. The office style is loose and casual, and the pace -- like Lime's trading -- is quick.

"The game is getting faster and faster," says Lime's CEO, a graduate of the University of Strathclyde in Glasgow, Scotland, who in 2000 co-founded this agency brokerage popular with statistical arbitrage hedge funds and other high-frequency traders. "People typing into a computer can't compete now with my type of clients. They can't compete with computers constantly firing off thousands of orders."

Like many players in the world of high-speed trading, Brown, 41, measures his business in milliseconds. Lime has developed technology that lets its clients place as many as 1,000 orders a second. For stat arbs, which use sophisticated computer programs to try to capitalize on short-term price anomalies in stocks and other securities, such systems are invaluable. As a result, Lime now handles about 4 billion shares a month for its 100 clients and is one of the three biggest brokers in New York Stock Exchange­listed companies traded on Nasdaq.

Lime is at the center of a furious new makeover in hedge fund trade executions, as U.S. exchanges ratchet up the speed of trading in response to technological and regulatory changes. As the floodgates are opened, the volume and speed of orders -- characterized by an exponential explosion in transaction data, which includes price quotes -- are swelling Wall Street trading networks. Some analysts expect daily U.S. trading volume to climb from 4 billion shares now to 6 billion by next year. The rush of orders coming from high-frequency traders is a bonanza for electronic execution providers like Lime; BATS Trading in Kansas City, Missouri; Automated Trading Desk in Mount Pleasant, South Carolina; NYSE's Arca (formerly Archipelago); and Nasdaq's INET.

One of the biggest electronic execution providers is New York­based Credit Suisse, which accounts for a reported 10 percent of U.S. equity trading volume. Most of that is done through its Advanced Execution Services (AES) unit, whose exotic algorithmic strategies are popular with hedge funds and other investors. About 40 percent of Credit Suisse's customers are said to be hedge funds.

The list of brainy hedge funds reaping the benefits of faster data access and trade execution is itself growing rapidly. Many of the biggest hedge fund firms, stat arbs like Chicago's Citadel Investment Group, New York­based D.E. Shaw Group and Renaissance Technologies Corp., headquartered in East Setauket, New York, get direct data feeds from the exchanges. But even small funds can access high-speed data from providers like Reuters and Bloomberg, which consolidate price quotes from myriad sources -- the NYSE, Nasdaq, regional exchanges and electronic communication networks (ECNs) -- and deliver it to investors in breathtaking time. James Conant, CFO of Archetype Risk Advisors, which runs several small quant hedge funds in New Haven, Connecticut, says his firm is looking to add equity trading strategies to take advantage of the changes.

"Low latency is important to us," he says, referring to faster computers.

Latency is the time it takes a trade to be processed and executed. For a hedge fund manager using computer programs to identify tiny mispricings, anything greater than submilliseconds can seem like forever. "To these players low latency is the oxygen in trading," says Louise Westerlind, a New York analyst for Boston-based research firm Celent. "Without it the data received will be obsolete and a potentially big profit can turn into a big loss."

Brown and his crew at Lime agree. "There's nothing much worse than trading on stale market data," he says. "You could lose your shirt."

As a result, firms like Lime have taken to moving their computers closer to the physical markets, in some cases right into the heart of the NYSE, to shave fractions of a second off transaction times. David Cummings, CEO of BATS Trading, says his firm has reduced transaction times substantially since it co-located its computers to New York and New Jersey from Kansas City. A typical trade now takes one one-thousandth of a second, down from 20 one-thousandths, he says, putting BATS in a statistical tie with Nasdaq's INET system for fastest execution. Lime's Brown won't have any of it. "I'd be happy to get into a huge bakeoff with everyone out there just to prove who's the fastest," he says. (Lime is an investor in BATS, as are Credit Suisse and other Wall Street firms.)

The prize is high-frequency hedge funds, especially the stat arbs. Statistical arbitrage, the province of mathematical traders, uses historical statistical patterns -- and expectations and assumptions -- to profit from short-term mispricings in securities. One of the most basic strategies is trading in pairs, say PepsiCo and Coca-Cola Co., buying one long while shorting the other. However, the strategies today tend to be much more complex, involving portfolios of stocks grouped together by sector.

"These stat arb traders may send in hundreds of limit orders at one time and then, seconds later, cancel and replace many of those with new limit orders," says Joseph Gawronski, president and COO of Rosenblatt Securities, an NYSE agency broker that is used by stat arbs.

Statistical arbitrage was popularized in the mid-1980s on the proprietary trading desk of Morgan Stanley. The strategy spread and attracted the brightest minds, including academic David Shaw, a former faculty member in the computer science department at Columbia University. His firm, D.E. Shaw, developed a reputation for exploiting inefficiencies in financial markets.

However, it was the advent of high-speed networks, especially the introduction of ECNs, that spurred the growth in statistical arbitrage and other high-frequency trading strategies. ECNs were sanctioned in 1997 by the Securities and Exchange Commission, when the agency, under pressure from irate investors, imposed new order-handling rules meant to prevent a repeat of the price-fixing scandals that had earlier convulsed Nasdaq. The rules were designed to protect limit orders, which reportedly had been routinely ignored by market makers. These sweeping changes were aimed at giving investors faster access and better prices -- and the same price-quote data that trading intermediaries had.

"In those days what you saw quoted on the screen was not necessarily what you got," says Matthew Andresen, who was president and CEO of Island ECN in the late 1990s and is now CEO of Citadel Execution Services (see box). "In today's marketplace what you see is what you get, and you get it right now."

As recently as 1997, Andresen adds, even ostensibly electronic exchanges like Nasdaq were, in effect, "call around" markets. Traders telephoned to negotiate. "When you're speaking on the phone, speed is not measured in milliseconds, it's measured in seconds and minutes," he says. "So being a little bit faster wasn't really an advantage."

When the equity markets moved from fractional to decimal pricing in 2001, the volume of price-quote data, already climbing, exploded, with more price points filling trading screens. At the same time, some hedge funds, suspicious that specialist firms might jump their orders -- and worried about tipping their hand in public -- turned to nontransparent systems, such as "dark pools" that put together buyers and sellers anonymously.

Although stock exchanges were ordered by Congress to compete more freely on price more than three decades ago, it wasn't until the SEC approved Regulation NMS in April 2005 that a potentially huge barrier to best execution was finally swept aside. Under Reg NMS, U.S. equity markets are required to provide investors with prices equal to or better than the best bid and offer, no matter where the price is quoted. At the same time, a shocked NYSE has been forced to modernize after seeing its market share siphoned away by rivals. Last year the NYSE rolled out an expensive, superfast Hybrid Market that integrates the Big Board's traditional specialist-based auction system with automated trading.

In the pre-Hybrid era many high-frequency hedge fund traders bypassed the NYSE, favoring instead ECNs and other faster electronic markets. "The order used to sit on the NYSE for 20 seconds or longer, which is an eternity in the eyes of an electronic trader," says former D.E. Shaw trader Daniel Mathisson, who moved to Credit Suisse eight years ago and now runs its Alternative Execution Services unit.

Today there are more opportunities to win in statistical arbitrage at the NYSE. The average turnaround time on many stocks traded at the exchange has shrunk recently by a stunning factor of 30 -- from about nine seconds, or 9,000 milliseconds -- to 300 milliseconds for Hybrid stocks.

"When you measure speed in milliseconds, the length of a cable between you and the exchange's computers really matters," says Robert Schwartz, a finance professor at Baruch College in New York and an expert on market structure. "If you are housed closer to the exchange, you have an advantage."

About five years ago, Lime began co-locating the computers that run its clients' algorithms and applications to its offices in TriBeCa, less than a mile from the Big Board. "My client could be sitting in a cornfield in Idaho," says Brown, casually. "As long as that client can access his trading machines, everything is fine. The algorithmic trading decisions are all done on my site."

This past summer, with trading volume rising, Lime ran out of computer space. So, Brown had a construction crew gut offices in nearby Jersey City, New Jersey, to house all of his clients' computers -- rows and rows of anonymous black boxes running the clients' programs.

A few hedge fund investors, though, are wary of the kind of high-speed trading described by Brown. David Saunders, founding partner of K2 Advisors, a $5.5 billion fund of hedge funds in Stamford, Connecticut, has shied away from stat arb funds, which are notoriously secretive. "If the investment methodology is clearly spelled out in an understandable way, so that I can manage my expectations about downside loss and upside gain, I think we can invest in them," says Saunders, a former trader at Tiger Management Corp. "We haven't seen that."

That kind of skepticism doesn't bother Brown, as he rides the high-frequency-trading wave. Sometimes, he won't hear from clients for weeks, but that's not a problem. "If they want to change their software, they can remotely log in and change it," he says. "They never need set foot in this office."

Inside Citadel Execution Services

On one excursion from Chicago to Hong Kong, Matthew Andresen of Citadel Investment Group logged some 7,800 miles in 22 hours. Hardly a record, but it was fast enough for Citadel's upbeat trader. Yet on a far shorter trip, 600 miles per hour doesn't come close to being satisfactory.

That trip is from Citadel's trading operation in Chicago to the New York Stock Exchange in lower Manhattan. Andresen's firm must electronically shuttle orders for execution to the exchange faster than any airplane -- practically at the speed of light, and light travels at more than 186,000 miles per second.

The trading at Citadel is propelled by supercomputers and fancy, high-speed routing technology. Still, in today's lightning-fast electronic markets, the firm's competitors in New York City have a slight time advantage -- measured in milliseconds -- because of their proximity to the Big Board. Delays cost money. That's why some hedge funds, like Citadel, have taken to installing, or "co-locating," computerized trading applications within sight -- or even inside the bowels -- of such exchanges as the NYSE and Nasdaq.

With $12.8 billion in assets and more than 1,000 employees, Citadel operates one of the biggest multistrategy hedge funds in the world. Founded by Kenneth Griffin in 1990, the firm recently became the first hedge fund to issue investment-grade bonds, selling $500 million in five-year notes. Andresen is president of broker-dealer Citadel Execution Services, the firm's market-maker affiliate, which trades about 150 million shares daily. The trading supervised by Andresen -- mostly silent, computerized executions -- is a vital component of Griffin's growth strategy.

"The superfast firms like Citadel Execution Services -- or any cutting-edge firm -- will want to be as fast as possible, so they are going to make their internal systems as fast as possible and have the best systems people and the best quantitative people behind how these are updated quickly," says the energetic Andresen, 36.

On a typical day, Citadel accounts for 3 percent of the daily volume on the NYSE and Nasdaq. Andresen, who also handles a pool of 500,000 options contracts daily, routes a large proportion of them to another New York destination, the all-electronic International Securities Exchange.

Every mile of cable wiring eliminated between Chicago and New York improves Citadel's transaction speed. It reduces latency -- the delay in the time it takes to deliver trade data to a computer and have it turned into an actual execution.

"You are limited by the telecommunications provider and by distance," says Andresen, a man who does not like to lose. (He was an alternate on the U.S. Olympic fencing team at the Atlanta summer games in 1996.) "So sending an order [directly] from Chicago to New York results in upwards of a 15-millisecond delay," he explains. "If you are competing, for example, with a New York market maker who has to run lines only a few miles downtown, you are automatically going to lose. They'll get the order executed faster."

Citadel's broker-dealer is a separate entity from the firm's hedge fund operation, but it is a significant part of the business activities of the firm. It eliminates transaction costs Citadel would otherwise have to pay to outside broker-dealers. And fees from customer executions contribute to the firm's bottom line. The liquidity pool at the broker-dealer also provides potential opportunities to internalize order flow -- that is, to match buy and sell orders in-house and conduct principal transactions as it commits capital on customer orders -- which can further lower costs for Citadel.

About half of the firm's assets are invested in equity strategies, such as convertible, merger and statistical arbitrage. Its $9.5 billion offshore flagship fund, Citadel Kensington, generated a net return of 16.95 percent in the first nine months of 2006, according to the prospectus for Citadel's bond offering. The $3.3 billion Citadel Wellington onshore fund returned 19.60 percent.

Citadel's sprawling trading parts are bound together by a high-speed technology platform that costs several hundred million dollars a year to maintain. One of the defining features is the record rate at which Citadel can receive and deliver data orders for itself and customers. Statistical arbitrageurs like Citadel use computer algorithms that obsessively track securities pricing data, sector by sector. The aim, simply, is to profit when one or more securities -- the roster can include hundreds -- diverges from a historical price range.

Of course, the strategy depends on having almost instant access to exchange data, which means stat arbs must find space near an exchange to plant their computer boxes. This phenomenon dates back to the late 1990s, when Andresen was running Island ECN, then an upstart computerized agency broker-dealer in New York. Island was the forerunner of several electronic communication networks promoted by Securities and Exchange Commission order-handling rules created to prevent price collusion by Nasdaq market makers. ECNs became popular among hedge funds and other customers because they were fast and anonymous, displayed quotes and protected limit orders. But Andresen says some Island customers in regional outposts started to complain.

"We didn't want people to win based on their geographic proximity to Island," he recalls. "We wanted to level the playing field, so we started doing co-locations for customers. And now this is something done today by all the markets, including Archipelago [now Arca] and Nasdaq."

At Citadel it has made all the difference. -- J.A.B.