Aug. 14 (Bloomberg) -- Knight Capital Group Inc.’s $440 million trading loss stemmed from an old set of computer software that was inadvertently reactivated when a new program was installed, according to two people briefed on the matter.
Once triggered on Aug. 1, the dormant system started multiplying stock trades by one thousand, according to the people, who spoke anonymously because the firm hasn’t commented publicly on what caused the error. Knight’s staff looked through eight sets of software before determining what happened, the people said.
Knight, based in Jersey City, New Jersey, hasn’t explained in detail what caused the trading losses, which depleted its capital and led to a $400 million rescue that ceded most of the company to a group of investors led by Jefferies Group Inc. The 45-minute delay in shutting down the malfunction has confused some securities professionals, who say that trading program can typically be disabled instantly.
“This software problem was an infrastructure problem,” Chairman and Chief Executive Officer Thomas Joyce, 57, said in an Aug. 2 interview with Bloomberg Television’s “Market Makers.” “It was more of a networking problem as opposed to using quantitative tools to trade.”
The company, whose market-making unit executes about 10 percent of U.S. share volume, will hire an outside adviser to investigate what led to the losses. Kara Fitzsimmons, a spokeswoman, said she couldn’t comment at this time.
Knight’s computers bombarded the market with unintended orders just after trading began on Aug. 1, causing volume to surge and prices to swing in dozens of securities. NYSE Euronext canceled trades that were 30 percent or more away from the price at the start of trading, a decision that applied to six securities out of 140 that were reviewed.
The company was updating software in preparation for an NYSE plan aimed at luring more individual investors to the exchange, Joyce said in the Aug. 2 interview, without offering details. The Big Board’s so-called retail liquidity program, designed to attract smaller investors by giving them superior prices, was being implemented that day.
Rules formalizing the treatment of erroneous trades were adopted amid criticism by investors after exchanges and the Financial Industry Regulatory Authority voided transactions totaling 5.6 million shares in the market crash of May 6, 2010. Regulators added guidelines governing when sales or purchases of stock could be canceled after market makers said confusion about which trades would stand prevented them from acting.
Getco LLC, an automated trading firm, Blackstone Group LP, brokerages Stifel Nicolaus & Co. and TD Ameritrade Holding Corp. and the investment banks Stephens Inc. and Jefferies paid $400 million on Aug. 6 for preferred stock that will convert into 267 million common shares. Knight, which closed at $10.33 the day before the trading error, ended yesterday’s session at $2.84.
The accidental trading at Knight follows the cancellation of an initial public offering by Bats Global Markets Inc. on March 23 and the May 18 IPO by Facebook Inc. on Nasdaq OMX Group Inc., which was marred by technology failures and delayed trade confirmations.
Knight’s episode underscores the challenges that Wall Street firms have faced updating and maintaining their computer systems. After Lehman Brothers Holdings Inc. collapsed in 2008, bankruptcy examiner Anton Valukas from law firm Jenner & Block LLP found that Lehman had “maintained a patchwork of over 2,600 software systems and applications.”
“The Examiner’s financial advisers ultimately requested access to 96 of the most relevant systems,” Valukas wrote in a report published in March 2010. “Many of Lehman’s systems were arcane, outdated or non-standard.”
Goldman Sachs Group Inc.’s technology staff accounted for 27 percent of its employee base in 2010, up from 16 percent in 2000, Chief Financial Officer David A. Viniar told investors at a conference in February 2011.
Copyright 2014 Bloomberg. All rights reserved. This material may not be published, broadcast, rewritten, or redistributed.