April 07, 2011

To help the buy-side combat the tactics of high frequency trading firms, earlier this week, Deutsche Bank's Autobahn equity business announced the launch of Stealth, a brand new liquidity-seeking algorithm for lit and dark venues.

The algo was built by Deutsche Bank's global team with all new code and infrastructure and is currently in beta testing with about a dozen clients.

Stealth is aimed at leveling the playing field for buy-side firms that are adjusting to today's electronic equity markets. "This tool is designed to let them exploit high frequency trading and low-latency trading techniques," says Jose Marques, Global Head of Equity Electronic Trading at Deutsche Bank, in an interview with Advanced Trading.

Stealth addresses fundamental changes in the equities markets, says Marques, who notes that electronic liquidity providers now account for more than 50 percent of the daily trading volume. "Specialists are no longer a human driven operation and upstairs market makers no longer exist," observes Marques. "A lot of the angst that the buy side has felt over the last couple of years has really come from this fundamental change. The buy-side trading community feels this keenly," says Marques. "You're sitting at the desk sending orders and the market behaves in a totally new way. It's very hard for people to interface with machines; they view problems differently and they react on fundamentally different time scales," he says.

From a conceptual view, Stealth is similar to typical opportunistic algos that source liquidity, but it's using built-in high frequency models to predict the next price change. "In the U.S. markets, we can actually predict the next price with an 80 percent accuracy rate," says Marques. The algo is looking at completion rates, at the evolution of the order book, and where events are happening and it's making a real-time forecast on the next tick. "But in the high frequency trading world, 80 percent accuracy is average performance," says Marques, adding, "As you shorten the time horizon, you get more accuracy. "

For example, if a trader is resting on the bid waiting for liquidity, and learns the next price change is forecast to go against it, and then the trader knows to cancel the bid and go across the spread, explains Marques. "The algo is using short-duration forecasts for placing hundreds of orders that span the life of institutional-sized executions. And, it's reducing the opportunity for the high frequency players," he asserts. However, the buy side trades as they always do. "Buy me a million share of IBM. I want to be between 10 and 15 percent of the market volume and I want to be done at 4:00 pm, and then the algo executes.