CHICAGO – Citadel Investment Group LLC’s high-frequency trading business began with a simple idea that was “easy to understand” and developed into computer models, according to Mikhail “Misha” Malyshev, a former Citadel employee who helped develop the program that made the trades.

The computer system’s speed, he said, was key to its success. “Number one will get the trade, number two will get nothing,” he said. “We were smaller than competitors, but the fastest in the world.”

Testimony last week in a lawsuit Citadel filed against former employees in its high-frequency unit shed light on the secretive business, which involves using complicated computer programs to predict small movements of share prices in order to make rapid, profitable trades.

Citadel filed the suit in the Chancery Division of Cook County, Ill., Circuit Court against Teza Technologies LLC, the group Mr. Malyshev and several other employees formed shortly after leaving Citadel early this year. Citadel alleged the group violated conditions of noncompete agreements that were signed. The defendants denied the allegations in court filings.

Jace Kohlmeier, a math whiz who worked for Mr. Malyshev at Citadel and then at Teza, testified on Friday about the types of problems he tackled in helping develop the high-frequency trading technique.

The group’s first task was to build a historical market-data system and develop trading signals called alpha models, according to a chart on Citadel’s business entered into evidence.

“Alpha models give you an idea, will something move up or move down,” Mr. Kohlmeier said. “Models just produce numbers, not decisions, so someone needs to have a strategy, someone who can turn those numbers into decisions.” Trading in the market gives traders “feedback” on the performance, which they used to improve the system, he said.

One task Mr. Kohlmeier discussed was trying to predict the bid change in shares of International Business Machines Corp. one second into the future, using such information as the size of a bid, the bid-ask spread and the imbalance between the two.

As the group produced returns that rapidly grew from a few million dollars in 2004 to more than $1 billion in 2008, some members of the group became concerned about compensation, and began “griping,” Mr. Kohlmeier said.

One source of discord was a report of better compensation from a former colleague who had joined a competitor.

Mr. Malyshev, who was born in a rural town in central Russia and came to the U.S. to study mathematics, eventually ending up at Princeton University, said the concept of using computer programs to predict share prices is “easy to understand,” at least at first. “You start with something simple, and eventually it becomes extremely sophisticated.” He compared it to a living organism. “It needs to be continuously updated to be successful,” he said.

After several months of development, the group had its first profitable day of trading using its system in July 2004. By 2008, Citadel’s returns from the group skyrocketed to about $1.15 billion, operating with around 55 people, Mr. Malyshev said.

A spokeswoman for Citadel declined to comment.