Monday, October 25, 2010

Coding Switch to Make Algorithmic Trading Even Faster

Coding Switch to Make Algorithmic Trading Even Faster

By DONNA KARDOS YESALAVICH
In algorithmic trading, where milliseconds can mean the difference between a profit or a loss, time has long stood in the way of getting the actual algorithms in place.

That lag time between conception and implementation could change soon, however. A standardized version of coding for automated trading programs is starting to gain traction, and that promises to bring traders new and revised algorithms faster.

Trading algorithms use software programs that make buy and sell decisions based on preset rules and opportunities that present themselves in the stock market. Their use has grown in recent years as trading firms have sought ways to capitalize on the ability of computers to more quickly spot and take advantage of market opportunities than a human could.

The coding, dubbed FIXatdl, for Financial Information eXchange Algorithmic Trading Definition Language, would help traders to get their hands on new or updated algorithms more quickly. That is a valuable prospect at a time when the industry is hungry for bringing fresh automated strategies into use.

"It's mostly a time-to-market issue for these algorithms," said Greg Malatestinic, senior technology analyst at Jordan & Jordan, a provider of regulatory and market-data compliance services.

Most programs require sophisticated coding that necessitates time-consuming changes at every step of the process, from the trading-system developers who would have to re-create algorithm screens for every new or altered algorithm to the trading desks where system upgrades would be needed to view the new screens. FIXatdl, however, makes time-sensitive algorithms a viable option that traders can profit from.

Mr. Malatestinic said an algorithm that takes weeks or even months in development, as many without FIXatdl often do, might be irrelevant by the time it is available. The FIXatdl standard would speed up the entire process, cutting the development time for algorithms to as short as a week or less.

The standardization aims to do for algorithmic trading what HyperText Markup Language, or HTML, did for the Web. Just as HTML enables Internet browsers to instantly read and show users all kinds of websites, the algorithm standard means trading firms will no longer need to install new technology each time they want to use a new algorithm.

In addition, reduced time to market means fewer man hours are needed to get algorithms into traders' hands, which ultimately helps cut costs.

"You're obviously spending a lot less resources and you have a quicker turnaround time," said Ary Khatchikian, president and chief technology officer at Portware, a developer of automated trading software that uses FIXatdl. It now takes Portware one to two weeks to do what took four to eight weeks prior to FIXatdl, Mr. Khatchikian said.

The idea of a coding standardization for trading algorithms isn't new. An industry group co-headed by Mr. Malatestinic has worked toward the goal for several years. But its first attempt at a standardized algorithm language, referred to as FIXatdl 1.0, was somewhat of a dud. That initial version, released in 2008, was stymied by glitches that limited its uses.

However, the FIX Protocol Algorithmic Trading Working Group, an industry group that developed and maintains the standard for algorithms, incorporated the industry's criticism of 1.0 into another version, called 1.1. That version, released in March, has generated a lot of buzz and increased adoption. The group doesn't directly profit from its work. It is composed of volunteers who work in the industry who decided a standard would enable them to work more efficiently.

"Version 1.0 probably got people engaged in thinking about it more," said Jim Campbell, Investment Technology Group's product manager for trading network ITG Net, while version 1.1 "puts us in a position universally to begin to move toward this standard."

ITG Net doesn't use FIXatdl but has plans to have some of the broker algorithms on its platform under the standard by early 2011.

"Really in the last quarter is when we feel like there's some critical mass that's been building up," he said. "Now that there's momentum here, we're certainly working with [the brokers] so we can adopt it."

Write to Donna Kardos Yesalavich at donna.yesalavich@dowjones.com

No comments: