High-Speed Traders Profit From Return of Loophole at CME
Five years ago, the world’s largest exchange operator vowed to fix a flaw in its systems that allowed high-speed traders to infer the direction of the futures market a fraction of a second before everyone else.
The problem arises from the two ways that CME distributes information about a trade. One is the private confirmation messages that the exchange sends to the buyer and seller in each transaction. The other is the public data feed that reports trades to everyone active at CME, a Chicago exchange where an average of 19 million contracts changed hands daily in January.
Sometimes, a firm will receive the private confirmation of its trade just before it is reported over CME’s data feed. During that delay—called a “latency”—an ultrafast firm can deduce that the market is about to move up or down, and quickly buy or sell to profit from that information, traders say.
A CME spokeswoman said the firm had “dramatically decreased the latency between public market data and private trade confirmations” since 2013, when the issue was exposed in The Wall Street Journal. But she acknowledged that private confirmations were still arriving first in some cases.
“Every market participant will experience some degree of variability,” she said in a statement. “In some instances, they may receive confirmations before the public market data feed and other times they will receive them after.”
CME was constantly working to improve its systems, she said.
The revelation that the flaw is still embedded in CME’s systems could lead to renewed scrutiny of the exchange operator. In 2014, during the uproar over high-frequency trading sparked by Michael Lewis’s book “Flash Boys,” CME Chief Executive Terrence Duffy was questioned about the issue on Capitol Hill.
Critics of high-frequency trading say exchanges like CME don’t have much incentive to fix such issues because their biggest customers are the ultrafast traders who benefit from them.
The delays to CME’s public data feed are measured in microseconds, or millionths of a second, and by all accounts are much smaller than they were five years ago. Still, the flaw can yield hundreds of millions of dollars in profit a year to traders, according to Quantlab Financial LLC, an electronic trading firm.
Data from Quantlab’s trading in 10-year Treasury futures—a popular interest-rate contract—suggest that CME quickly fixed the latency problem in 2013, but for reasons that are unclear, it crept back into the exchange’s systems two years ago. By December, the median private-over-public advantage that Quantlab saw was around 100 microseconds. In about 10% of the firm’s trades, the delay was well over 2,000 microseconds.
Such latencies can vary widely. Still, people at several other high-speed trading firms confirmed they regularly saw similar delays in other CME markets, including futures tied to the S&P 500 and crude oil. One of the people said his firm often saw delays of more than 200 microseconds in oil futures.
“This creates a special club of firms that benefit from the information asymmetry,” said John Michael Huth, chief operating officer of Houston-based Quantlab.
There are various ways to potentially exploit the flaw. Typically, such strategies involve so-called canary orders—small buy or sell orders, for one or two contracts, used to detect large trades that could move the market.
For instance, if oil futures can be bought for $60.01 and sold for $60, a trader could place a small order to buy at $60, which would join a queue of similar buy orders at CME. If the trader gets a message saying his or her buy order was filled, that could signal that a large seller is at work and the price is about to tick down to $59.99. The trader could quickly sell at $60 to take advantage of the expected move.
Alternatively, a canary-order strategy could be used in markets that closely track CME’s futures, such as the SPDR S&P 500 Trust, the popular exchange-traded fund best known by its ticker, SPY. If a trader detects that CME’s S&P 500 futures are about to tick up, he or she could quickly buy SPY before others learn about the futures move.
Mr. Huth said such practices harm ordinary investors, who rarely trade futures themselves but often hold shares in mutual funds and other vehicles that use futures to protect against market volatility. He acknowledged that Quantlab could exploit the CME defect itself but had decided to publicize its concerns instead.
“We are trying to effect change so that markets are on a level playing field,” he said.
Other traders rejected the idea that investors were being harmed. Instead, they said the main impact of CME’s flaw was that it provided an edge to a few large, sophisticated high-speed traders at the expense of smaller rivals.
Still others suggested that it was appropriate for CME to notify traders that their orders had been filled before notifying the public. After all, the only thing traders are learning is whether their own orders have been completed, these people said.
“Reasonable people can disagree about whether it’s more ‘fair’ for trade information to be sent on the public data feed before or after traders receive their fill confirmations,” said Kipp Rogers, a former trader who is now a blogger and researcher.
Other exchanges have grappled with the same issue. Eurex, the derivatives arm of Deutsche Boerse AG , says its public feed now beats its private confirmation messages 99.9% of the time. Private messages were generally faster before 2012, when the German exchange overhauled its systems and prioritized the public feed, said Randolf Roth, a Eurex board member.
That decision sparked an outcry from some customers, Mr. Roth said. “Not everybody was happy about it,” he recalled. “A lot of U.S. firms in particular wanted a speedy model.”
Write to Alexander Osipovich at firstname.lastname@example.org