The most important rules governing high-speed electronic trading weren’t written by market regulators, New York’s attorney general or a computer whiz. They were formulated a century ago when Albert Einstein figured out nothing could travel faster than light.
High-frequency traders, seeking to exploit money-making opportunities first, are moving information at velocities approaching Einstein’s barrier. The competition has become so extreme that earlier this month one vendor published a press release to boast it had shaved 12 millionths of a second off the transmission time between New Jersey and Illinois.
Caught up in a race that must end in a tie, firms that buy and sell securities in seconds while jumping across trading venues and assets are looking into new strategies that actually involve slowing down. The laws of physics are running into the laws of the state of New York, where Attorney General Eric Schneiderman said this week he wants to rein in the fastest traders.
“No one’s messing with Einstein,” Ari Rubenstein, chief executive officer of Global Trading Systems LLC, said in an interview on Feb. 28, before Schneiderman’s probe was announced. “Our absolute speeds might be increasing, but our relative speeds are approximately the same. Investments right now in technology are more about detente than arms race.”
In an industry where faster has always meant better, HFT firms are now reaping fewer rewards from incremental speed gains as brokerages improve their technology to defend themselves. At the same time, Schneiderman’s investigation targets the “unfair” services that allow them to pursue many of their strategies. Speed traders are also being pressured by a less-volatile stock market, which makes it harder to profit from very short-term techniques.
As high-frequency trading gets harder commercially, legally and scientifically, use of super-fast strategies has expanded beyond HFT, according to a Feb. 11 report by analysts Anshuman Jaswal and Muralidhar Dasar at research firm Celent.
Last year Morgan Stanley, whose equities trading unit accounted for more than 10 percent of average daily volume in the U.S., completed an overhaul of its infrastructure so that it could save clients fractions of a millisecond. Brokerage firm Investment Technology Group Inc. spends more than $40 million a year on capital expenditures, most of it on technology to help clients keep up on speed, said spokesman J.T. Farley.
Transmission time is one of many areas ITG’s clients wanted to improve in an “industry with a lot of firms with big balance sheets, a lot of capability and who are highly sophisticated in terms of artificial intelligence and algorithms,” CEO Robert Gasser said in a Feb. 24 interview.
The decline in price swings in the stock market is also a challenge for speed traders. Although the details on how HFT firms make their money are closely guarded secrets, many strategies revolve around old-fashioned market making -- simultaneously providing both bids to buy and offers to sell.
Rubenstein’s GTS operates as an automated market maker handling about 3.5 percent of U.S. stock-market volume and between 4 and 5 percent of U.S. Treasury futures trading.
HFT firms profit more in volatile markets because prices move faster and to a greater extent, enabling them to buy lower and sell higher.
The Standard & Poor’s 500 Index has swung an average of about 0.9 percent between its daily high and low in the past 12 months, compared with 2.7 percent in 2008. The Chicago Board Options Exchange Volatility Index, or VIX, a measure of U.S. stock volatility, rose above 80 in 2008 and traded at an average of 32.69 that year. This year, it’s averaged about 14.86.
The calmer markets have dented earnings in the industry. HFT profits fell to as little as $810 million in 2012 from $4.9 billion in 2009, according to the latest estimates from Rosenblatt Securities Inc. HFT was involved in about half of volume in 2012, compared with 66 percent in 2009, according to the firm.
The competition for speed was a result of common trading rules at exchanges and alternative venues. In the system known as “price-time priority,” equivalent-priced orders to buy or sell are executed according to when they reach the exchange’s trade-matching computers.
So firms made large investments in reducing their systems’ latency, or the time it takes to make the electronic connections required to trade in dozens of data centers around the world. By 2011, performance-monitoring firm Corvil Ltd. was offering products that could track latency down to the nanosecond, or one-billionth of a second.
“Technology allows for the automation and scale that greatly lowers costs for investors,” Peter Nabicht, senior adviser to the Modern Markets Initiative trade group and former chief technology officer at HFT firm Allston Trading, said in an e-mail. “Professional traders do not compete with investors, they compete with each other to provide the liquidity and fair prices that today’s investors enjoy, investing in low latency and cutting edge technology to do so.”
Some of the practices used to cut latency, such as co-location, where a firm places its servers in data centers next to exchanges’ systems, are now in Schneiderman’s sights. New York’s top law enforcer said that with high-frequency firms, “the real problem is they are able to thrive because they get special, early access to information that isn’t available to the rest of the market.”
Schneiderman is right to investigate the “plumbing” of the market to determine if firms are getting special advantages from innovations such as proprietary data feeds, sophisticated order types or payments for order flow, according to Joe Saluzzi, co-head of equity trading at Themis Trading LLC and a frequent critic of market structure.
HFT firms are “out there trying to beat each other to death, essentially is what they’re saying,” Saluzzi said in a phone interview. “They’re racing against each other, but in the end we always know who the loser is,” he said. “The loser’s always going to be a non-HFT.”
While advances in market structure have improved prices for investors, quickened trade executions and reduced fees, the risks from the complexity are amplified by the increase in speed, Gary Cohn, president and chief operating officer of Goldman Sachs Group Inc., said today in an opinion piece for the Wall Street Journal.
Cohn said the equity market needs a stronger “safety net of controls” to block problematic orders from entering the market. Incentives should be created to reduce excessive instability and mitigate stress caused by trading techniques that cause high volume, he said. Public market data should be disseminated to all participants simulutaneously, he added.
The scrutiny from regulators has firms balancing the costs of improving speeds with rising expenses to comply with stricter rules, according to Celent. Decisions by investment firms about “squeezing the network for incremental latency reductions require greater deliberation now when compared to two years ago,” the firm’s report said.
GTS’s focus on latency has grown to include work on more sophisticated real-time compliance and risk checks, as well as ways to perform “disaster recovery” without halting trading. The firm has also needed to improve speed when it comes to responding to regulators.
Rubenstein, who started his career in cotton and orange-juice trading pits alongside some of the same people used as extras in the 1983 Eddie Murphy comedy “Trading Places,” said regulators now contact his team “all the time,” compared with once or twice a year a decade ago.
“The regulators in the last few years have gotten orders of magnitude more sophisticated,” he said. “And in real time they’re constantly, ‘Hey can you explain why you sent these orders.’ They’re sending us letters, they’re calling us. We need to extract data efficiently, get it to them and it has to also make sense. You have to be able to interact with regulators efficiently or you will constantly be crippled by those interactions.”
One of the most high-profile measures of latency is the time it takes to move information between data centers in northern New Jersey and machines in Illinois.
A quick connection creates opportunities to profit from fleeting discrepancies in assets traded in the different markets. For example, prices for S&P 500 Index futures may move in Chicago before exchange-traded funds tracking the index move in New York, or vice versa, providing brief opportunities to arbitrage the difference.
The route’s importance to HFT firms became clear in 2010, after Spread Networks announced it had spent three years boring through 825 miles (1,328 kilometers) of terrain in secret to lay an underground cable between Chicago and New Jersey. The wire cut three milliseconds, or three-thousandths of a second, off the previously fastest route.
The focus has now shifted to faster wireless microwave connections.
A firm called McKay Brothers boasts of making the round trip via microwave in about eight milliseconds. The market for the service is expanding beyond first adopters and “will be mainstream very soon,” co-founder Stephane Tyc said in a March 3 statement announcing the firm’s latest transmission times.
By cutting the time between Carteret, New Jersey, and Aurora, Illinois, to 8.12 milliseconds, the network “is rapidly approaching the physical limits,” co-founder Bob Meade said in the statement.
Returns on investment are diminishing as firms now look to shave nanoseconds instead of microseconds or milliseconds off of their latency. Computer-driven trades can now be executed in about 300 microseconds, according to one study. At that speed more than 1,000 trades can be made in the blink of a human eye, which lasts 400 milliseconds.
High-frequency strategies in most cases are “as fast as they need to be in order to be successful,” Modern Markets Initiative’sNabicht said in a March 7 interview.
“We are getting to a point where there is less edge in getting faster because it’s smaller and smaller increments in speed you’re shaving off, and there’s more edge really in the smarts, in the strategies,” he said.
Sophisticated investors, such as statistical-arbitrage funds that use mathematical models to identify mispriced securities or predict future moves, have also ramped up the speeds at which they deploy their strategies, eroding HFT firms’ primary advantage. That’s forced the speed traders to adopt more complex techniques with longer holding times, according to Manoj Narang, CEO of Tradeworx Inc., in Red Bank, New Jersey.
“Stat-arb firms are adopting a lot of the high-frequency traders’ techniques for executing their trades to avoid leaving pennies on the table for the high-frequency guys,” he said in a phone interview on Feb. 13. “That’s putting pressure on the high-frequency guys’ business because the stat-arb guys are basically eating their lunch. And in response, the high-frequency guys have no choice but to start being smarter about their positioning. That manifests itself in longer holding periods.”
Tradeworx is holding positions for 10 to 11 minutes, double the average in 2010, according Narang.
The move into trades based on more predictive analysis is a “natural expansion,” GTS’s Rubenstein said. His firm has been talking to strategists from statistical-arbitrage hedge funds that turn over positions more slowly than HFT shops, in time periods ranging from five to 20 days.
The shift is a “classic diversification,” according to Michael Kearns, a professor at the University of Pennsylvania who has run quantitative trading teams on Wall Street. The fast response time of HFT trading usually means firms can only place small bets on each trade, while longer time horizons for stat-arb firms allow them to put more money to work, he said.
“Even though you can put less money to work in HFT, you can generate much higher returns, so it’s also a natural for traditional quant stat-arb groups to start to introduce HFT components to their portfolio,” he said in a telephone interview Feb. 25. “If you can do both of them reasonably well, you’ll get something that’s more than the sum of its parts.”
To contact the editors responsible for this story: Nick Baker at email@example.com Jeremy Herron