Climbing the Algorithmic Learning Curve
Institutions seeking anonymous access to U.S. equity markets are flocking to algorithmic trading - a computerized strategy that slices and dices large orders into smaller pieces to avoid market impact. With the potential for reducing transaction costs and measuring returns against a chosen benchmark, algorithmic trading has become one of the hottest trends on buy-side trading desks over the past two years.
Nahn Bui, head of equity trading at First Quadrant, a quantitative investment-management firm based in Pasadena, Calif., with $16 billion in assets, gets frustrated when she can't execute a 10,000-share order. "The market is so thin now, it doesn't make much sense for me to show a block of stock to the floor [of the New York Stock Exchange] because they'll move the stock on me," says Bui, who resorts to algorithmic trading because it offers more anonymity over traditional trading methods. "The spread is so small, the transparency is gone. I rely on the rule-based trading to alleviate some of my pain."
Instead of placing a 10,000-share order, an algorithmic trading strategy will, for example, push 300 shares out every 30 seconds and incrementally feed small amounts into the market over the course of several hours or the entire day. The time frame depends on the trader's objective, how passive or aggressive he or she wants to be and constraints such as size, price and order type, as well as liquidity and volatility of the stock and industry group.
According to a survey of 52 buy-side institutional traders conducted by The Tabb Group, 60 percent of the firms use "model-driven execution vehicles." Though most buy-side firms are in the testing phase, says the Tabb report, responses show that buy-side firms are using automated trading for 5 percent of their order flow, and the firm predicts that that will increase to 13 percent by 2006.
So what is algorithmic trading and why is it catching on? How are buy-side and sell-side institutions using it? And what impact is it going to have on trading desks?
Algorithmic trading "is a way to codify a trader's execution strategy," says Jana Hale, global head of algorithmictrading at Goldman Sachs in New York. "It takes how traders think and strategize about their flows and then implements that in an electronic and systematic way."
Traders use the models to obtain the returns of certain benchmarks - the most common of which is volume weighted average price (VWAP). But brokers offer strategies beyond the simple VWAP. "They've gotten far more complex and far more advanced than a simple volume weighted price model," says John Wheeler, vice president and director of U.S. equity trading at American Century in Kansas City. Strategies offered by brokers include: time weighted average price (TWAP), the previous night's close and implementation shortfall - a model that weighs the urgency of executing a trade against the risk of moving the stock.
"The whole game is about balancing time versus impact," while staying within the client's constraints for price, time and volume, explains Dan Mathisson, global head of advanced execution services (AES) at Credit Suisse First Boston (CSFB). There is a trade-off between "stretching out [the trade] versus how much impact it will incur, and picking the right times to be executing the stocks," he adds.
Tweet This