Friday, January 06, 2006

Black box trading thrives in FX market

By Gertrude Chavez-Dreyfuss
(c) 2006 Reuters Limited

NEW YORK, Jan 5 (Reuters) - The deeply liquid and often volatile global currency market has become fertile ground for algorithmic or "black box" trading models, an arcane bunch of computer systems designed to execute trades that are unswayed by human biases.

Black box models, which became popular in the stock market in the 1980s, utilize sophisticated mathematical models to make currency trade decisions that some users claim are quicker than the blink of an eye.

One of the things they can do is quantify relationships among different currencies, traders say, and using historical trends as a guide, aim to capture discrepancies in prices across various trading platforms.

With about 20-25 percent of global currency managers using some form of black box model to execute orders, analysts say algorithmics is a growing space in the market.

The surge in daily currency volumes and the introduction of electronic trading platforms have spawned a boom in algorithmic models, according to analysts.

The Chicago Mercantile Exchange (CME), for one, has attributed the 49 percent jump in currency futures trading last year to what it calls "high-velocity" algorithmic traders.

"Black boxes work in foreign exchange because there is a lot of model uncertainty in the market," said John Taylor, chairman and chief investment officer at FX Concepts, a currency hedge fund in New York with about $11.3 billion in assets under management.

About 80 percent of FX Concepts' currency transactions are model-driven, while 20 percent are trades that use some discretion, according to Taylor.

"It's pure mathematics that drives the markets," he noted.

The impact of algorithmic models was best illustrated in the dollar's price action against the yen last month, traders said.

When dollar/yen , which had risen for most of 2005, hit around 122 in December, model accounts sold the pair because their computer signals were screaming a "big sell" on the dollar. As a result, dollar/yen fell to around 115 yen, which had nothing to do with fundamentals, some said, resulting in fat profits for some model accounts.

"Fundamental announcements are important information packets but have to be balanced against market expectations and the daily activity of the markets," said Mark Rzepczynski, president and chief investment officer at John W. Henry, in his monthly market commentary.

J.W. Henry Co., an alternative asset manager based in Florida with about $3 billion in managed assets, is also a huge model-driven fund, using analysis derived from market prices.

Market participants also cited the success of mathematician Jim Simons' $5-billion Medallion Fund, one of the most compelling stories on algorithmic trading. Simons, who according to hedge fund managers prefers to stay low-key, is the founder and president of investment management firm Renaissance Technologies in New York.

Simons, through a Renaissance spokesperson, has declined requests for an interview.

Market insiders and news reports say his Medallion Fund, which invests in currencies and other asset classes, has averaged about 35 percent in annual returns, net of fees since 1989.

His hedge fund employs top academic experts, the majority of whom have doctorate degrees in the hard sciences, according to other hedge fund managers. Their job is to pore over market data to unravel statistical correlations that could forecast price movements of commodities, currencies, and stocks.


Still despite their proven knack for making profits, black box trading is highly-competitive and not for the faint of heart.

Analysts say algorithmic models are constantly changing and can sometimes be replicated so traders need to continuously update their systems to ensure that they're always ahead of the competition.

Black box trading has also raised some crucial issues, such as the potential for models to miscalculate and cause a meltdown in the market.

Skeptics are quick point out that Nobel laureates had written the trading algorithms for Long-Term Capital Management (LTCM), which collapsed in 1998. LTCM eventually lost $4 billion after a Russian debt default.

"The danger in relying on model signals is that they're not very fluid. No black box can jump in and out of the market. Also in a trendless market, models get killed," said Erste Bank trader Joe Francomano, who uses a blend of technical and fundamental analysis in currency trading.

At the end of the day, most skeptics say there is no substitute for good intuition and sound judgment in making currency trades. "Even the black box is programmed by somebody who writes the codes. You at least must have some discretion in doing that," said Francomano.

© 2006 Dow Jones Reuters Business Interactive LLC (trading as Factiva). All rights reserved.

No comments: