Dark Algorithms Solving Fragmentation Issues
The reports are in: Buy-side firms have conquered their fear of the dark and are diving into dark pools of liquidity armed with a new battery of algorithms from their brokers.
Although there always has been some degree of hidden liquidity in the marketplace -- in the form of reserve or resting orders on ECNs, for example -- the phenomenon of tapping dark pools to try to find hidden order flow is now a major factor in decision-making on buy-side trading desks. A survey published in December 2006 by Financial Insights noted that in the space of a year, the percentage of firms using algorithms to access dark pools had risen to 60 percent, up from 32.5 percent the previous year.
Meanwhile, use of public exchanges, such as the New York Stock Exchange (NYSE) and Nasdaq, as well as traditional crossing networks, has declined. And while none of the dark networks has crossed the regulatory threshold requiring any trading environment accumulating more than 5 percent of U.S. equity volume to publish quotes to the public market, dark pools have become a significant enough destination that brokers now are investing considerable resources in either building their own, creating algorithms explicitly to reach multiple dark pools or both.
Goldman Sachs, for example, is in the process of reengineering its entire algorithmic suite to support dark-pool navigation. By the end of this summer, Goldman's volume-weighted average price (VWAP), participation, portfolio-trading and implementation-shortfall strategies will be dark-pool-compliant, along with Sonar, the liquidity-seeking algorithm that already accesses dark pools, according to Douglas Borden, head of the U.S. algorithmic trading desk and algorithm development at Goldman Sachs.
Accessing Dark Liquidity
"Clients are getting more comfortable with the liquidity that is available," says Borden. "The buy side's initial concerns about information leakage have been resolved, as most platforms have built in some sort of protection against gaming. The bigger question now is whether the dark liquidity is accessible."
Goldman's reengineering of its existing algorithms, as well as the deployment of Sonar, involves applying the methodology of a liquidity-seeking algorithm -- which is intended to mimic the investigative behavior of traders on the desk -- to the objectives of the existing strategies. For example, in a VWAP algorithm, the expected volume profile for a stock is assumed, given historical data, and a trading schedule is established. The order generally is traded in very small lots across multiple venues to sample the greatest range of offers, as well as to achieve greater anonymity.
If there is a need to use dark liquidity, the algorithm must collect statistics about how much liquidity is available in each dark book. Unlike in the public markets, real-time data and quote advertisements are not available, so a small, experimental trade must be conducted. If an execution occurs, it is usually safe to assume there may be more interest behind the trial order.
"You assume a certain number of executions in those venues and route some of your child orders to some of the dark pools," Borden explains. "As the trade goes on, you need to update how many fills you are getting to stay on the trading schedule."
To aid in the battle, and to enter the burgeoning market, Goldman Sachs has created SIGMA X, an internal dark pool, as a companion to Sonar. SIGMA X recently crossed 100 million shares in a day, says Borden.
Clients tend to enter SIGMA X in one of two ways: as a first stop on the way to the public market for market orders and marketable limit orders, and as a possible destination for limit orders, according to Borden.
For orders that require immediate execution, Goldman first passes the orders through SIGMA X and then on to the public markets if they cannot be filled completely in the dark pool. For limit orders -- whose limitation is a certain price and/or time threshold in the market beyond which trading is to be terminated -- the best price is sought methodically in all pools, dark and public, with no preference given to SIGMA X, says Borden. He estimates that about 10 percent of Goldman's total executions in U.S. equities occurs in dark pools.
The plan is similar at Banc of America Securities, which is planning the launch of FUSOR, an algorithm that searches all available venues for liquidity using a conceptually similar method to Goldman's: Dark venues are "pinged" with small experimental orders before the algorithm goes in for the kill. The idea is less to make a big noise about dark pools than it is to obfuscate the differences between liquidity pools and make the trading experience seamless, explains Bill Harts, managing director, Banc of America Securities (BAS).
Tweet This