How High-Speed Traders Outraced the Profits

Traders work in the Rosenblatt Securities booth on the floor of the New York Stock Exchange Photograph by Scott Eells/Bloomberg

For the last few years, the stock market has been a race among high-speed traders. The game was simple: The fastest traders won the most profits. Since speed was all that mattered, trading got exponentially faster. In the last five years, the time it takes to trade a stock went from being measured in milliseconds to microseconds. The fastest firms can now execute a trade in under 10 microseconds. It takes 350,000 microseconds just to blink your eye.

Now it appears the advantages of speed are starting to dissipate, and being the fastest trader isn’t worth what it once was. High-frequency trading profits are expected to fall 35 percent this year, 74 percent below their peak in 2009. In an ironic twist, high-frequency traders have gotten so fast, they seem to have outrun their own profitability.

Though a lot of trading has migrated off exchanges into private trading venues known as dark pools, high-speed traders still operate largely on exchanges. But while the traders have continued to get faster, pouring millions of dollars each year into improving their execution times through better software and equipment, the exchanges haven’t kept pace.

As a result, the firms that show up first in line to execute a trade have to wait for the exchange to catch up, and in a matter of microseconds, their speed advantage disappears. “The equipment of many stock exchanges recognizes a trade only every 30 to 40 microseconds,” says Rick Cooper, a finance professor at the Illinois Institute of Technology, who built trading algorithms and quant models for State Street Global Advisors in the 1990s.

Today the firm that shows up first with an offer to fill a buy or sell order is no longer guaranteed to win the trade. “The first guy in line now has maybe a 60 percent chance of winning,” says Cooper. “It’s random.” Those are still decent odds, but in the 40 percent of cases when the first guy in line doesn’t win the trade, all those millions of sunk costs spent on staying at the head of the trading frenzy end up wasted. “That kills profits,” he says. A spokesman for the New York Stock Exchange declined to comment on the exchange’s speed. A spokesman for the Nasdaq OMX Group said that the exchange can receive a trade and then send an acknowledgement back to the trader in under 40 microseconds.

On top of that, trading volumes are way down, and so is volatility. High-frequency traders do best when volumes are high and the market is choppy, with prices moving rapidly up and down. A low-volume, placid stock market is the worst kind for speed traders, making it even harder for them to eke out profits.

To HFT critics, this turnabout may come as just desserts for a phenomenon viewed by many as the market’s bogeyman, the computer-driven specter that spooked traditional long-term investors out of the market for fear that it had been overrun by swarms of predatory trading algorithms. Accurate or not, high-frequency traders are now clawing at each other for smaller bits of a shrinking pie.

The changed landscape isn’t a death knell for HFT, but it does signal an inflection point in their evolution process. “Sophistication and adaptation is where the future lies,” says John Bates, chief technology officer for Progress Software, a Bedford (Mass.)-based firm that builds algorithms for trading firms. “The strategies of the future won’t just be based on speed but on smart, new forms of arbitrage and market analysis. It’s no longer good enough to be the fastest.”

In other words: The cheetahs need to grow bigger brains if they’re going to survive. That means melding speed-based quantitative trading methods with more traditional qualitative ones. Instead of building a trading algorithm based entirely on structured data like bid and ask volume and pricing data across a handful of exchanges, speed traders are increasingly feeding their algorithms with qualitative data like news stories and earnings reports, or maybe a speech by Ben Bernanke.

Writing an algorithm that can effectively interpret these new sorts of data is a much higher bar than just crunching numbers and scouring markets for tiny price discrepancies. The more complex the strategy, the more it requires tweaks.

“Strategies like that become increasingly difficult to build and maintain,” says Ben Van Vliet, a finance professor at the Illinois Institute of Technology, who has co-authored papers on the risks of speed trading with Cooper, his colleague at the Chicago school. “Trading strategies that used to work for six months may now work only for six weeks, or six days.”