Never before has the simPle act of buying and selling stock seemed so mind-numbingly complex. As electronic systems supplant old-fashioned trading floors, today’s equity traders face a dizzying array of options for executing their orders, from floor brokers to computerized algorithms.
Pipeline Trading Systems, a three-year-old electronic market, is trying to bring some order to the chaos. One of several so-called dark liquidity pools giving market participants a place to trade blocks anonymously, Pipeline averages 40,000 shares per trade, compared with an average of just a few hundred shares for the New York Stock Exchange. Traders have flocked to the system, which handles some 32 million shares daily, up threefold from two years ago.
But most trading still occurs away from Pipeline and other dark pools — on markets like the NYSE and the Nasdaq Stock Market, which require users to reveal more information, including the prices at which they want to trade. Pipeline wants to capture some of this trading, about half of which is done using algorithms that chop big orders into hundreds of smaller ones, executing them over time in multiple places at the best available prices. Algorithms have so proliferated — virtually every brokerage firm of consequence offers several tailored to various trading styles and strategies — that traders are often befuddled about which ones to use and whether they work.
Pipeline hopes to capitalize on that confusion with a new system it calls an algorithm switching engine. The firm, partnering with several brokerages it declines to name, will let traders using Pipeline’s dark pool make a portion of their orders available for execution through these firms’ algorithms. Pipeline’s system analyzes these orders using 50 trading characteristics — such as the stock symbol, number of shares and time of day — to predict market behavior, while continuously selecting the optimal execution style for each order at a given point in time. Depending on market conditions, the engine may use several algorithms over an hour or two to execute the part of an order that isn’t filled in the block book. Pipeline built the system with Adaptive Technologies, a predictive analytics company co-founded by Pipeline research head Henri Waelbroeck and Santa Fe Institute professor J. Doyne Farmer, co-founder of research firm Prediction Co.
“The fact is that 95 percent of the shares our customers have to trade are not getting done in the block market, and there’s an opportunity cost there,” says Pipeline’s president, Fred Federspiel. A former nuclear physicist for the Los Alamos National Laboratory in New Mexico, Federspiel learned about trading as a salesman for technology consulting firm Bios Group and co-founded Pipeline as a division of Nasdaq in 1999. (Nasdaq has since spun off Pipeline but retains a minority stake in the 60-person firm, whose chairman is ex-Nasdaq president Alfred Berkeley III.) “Our customers have been asking us to find a way to marry block discovery with the rest of the market,” says Federspiel.
Pipeline will keep a portion of the commissions paid to the brokerages whose algorithms are part of the switching engine. About a dozen broker algorithms were accessible through the system when it launched last month, and Pipeline is in talks with several brokerages about adding more, says Federspiel.
Pipeline’s initiative is part of a trend toward merging the liquidity displayed in public markets, such as the NYSE and Nasdaq, with the dark liquidity in crossing networks, like Pipeline’s block book. Liquidnet Holdings, a Pipeline competitor, last year introduced a system called H2O that exposes shares in its crossing network to algorithmic flow from brokerage firms. More recently, Liquidnet acquired algorithmic shop Miletus Trading and began working on ways to integrate that firm’s business with its own core crossing network. And brokerages like Credit Suisse have negotiated deals with dark pools to include them as execution destinations for certain of their algorithms.
Pipeline is the only firm that claims the ability to help clients pick which algorithms will work best for specific trades. There’s plenty of potential there, but also the risk of alienating customers by making the wrong recommendations.
“It’s a great idea,” says a competitor of Pipeline’s new venture. “But you’d better be sure you’re giving people the right advice. I wouldn’t want to have people coming to me saying they lost all kinds of money because I told them to use the wrong algo.”
