HispanicBusiness

Some algorithms are relatively simple, and target a stock's average daily price. Others are more complex, trading according to market volume and certain times of the day. "Most algorithms trade against a benchmark like market indices or certain rules created within the algorithm," Mr. Lee explains. "There are many benchmarks, and traders can theoretically accommodate any risk level."

Large Wall Street trading firms have used algorithms internally for years. But the concept has changed drastically over the past few years, for several reasons:

• In 2001, Wall Street changed from valuing stocks in sixteenths to pennies, increasing the number of price points for each dollar, reducing trading margins and profits for brokers.

• Growing government scrutiny of Wall Street and increasingly savvy individual investors made trading equities less profitable and more complicated than ever.

• Most importantly, electronic communications networks (ECNs) grew, allowing brokers to trade securities faster and more efficiently through exchanges worldwide. Large trading firms that had used algorithms internally began to use them for client trades.

Stocks with large capitalizations account for the bulk of algorithm-traded securities. But firms are starting to use algorithms for government securities, futures, and options. Mutual funds and pension fund managers are also beginning to make the move.

Algorithm-based trades will account for 40 percent of U.S. equities trading by 2008, compared with 25 percent currently, according to Aite Group.

Over the past few years, big financial institutions such as Credit Suisse First Boston, Goldman Sachs, Banc of America, and Lehman Brothers have built sizeable departments that develop algorithms and improve them. Meanwhile, smaller competitors also emerged. "We have seen a jump in the number of smaller firms that are pure algorithm providers run by people from large firms and hedge funds," says Mr. Lee.

Customized Vs. "White Label"

Most of the firms provide off-the-shelf "white label" algorithms that aren't tailored to individual institutional clients, says Mr. Lee. But ATS employs a team of engineers who create customized algorithms.

Customized algorithms are more expensive than the white label variety. "Large firms spend millions of dollars on customized algorithms. For a mid-sized firm like ours, the best choice was to outsource it and form the joint venture," says Mr. Ramirez. With the ATS-customized algorithms, Ramirez Trading Solutions can position itself to potential clients as a shop that offers its own unique services that can do the following:

• Assess the potential risks and costs of clients' trades and trading strategies,

• Analyze performance against specific preset benchmarks, and

• Keep trades and strategy confidential by trading in large blocks.

Still, algorithm-based trading poses some risks, Mr. Lee says. At many institutional brokers, he explains, client orders pass through the firm's proprietary trading desk before they are executed algorithmically. The process can raise trading costs, delay trades, and create a potential conflict of interest between a brokerage's prime trading desk and its algorithmic trading operations. For example, the prime desk could pass a huge sell order on to the algorithmic trading unit but use the information to benefit an order from a non-algorithm trading client ["trade against" the algorithmic order]. Also, firms that don't use algorithm trading expertly can unknowingly create patterns that other traders recognize.

None of that would happen with RTS, says Mr. Rodriguez. They deal in superior expertise, and "have structured it so that we will never trade against customer orders."

Service is Human

Algorithms can significantly reduce overhead costs for brokerages that use the technology expertly. "Algorithm trading requires sophistication that not all small and mid-sized firms possess, but Ramirez & Co. has that sophistication," says Mr. Rodriguez.

Algorithms can replace human traders, but Ramirez & Co. has no such plans, he adds. His reasoning? Algorithms are no better than the people who use and program them, and personal relationships with clients will always be important.