After losing $440 million in 45 minutes of trading Wednesday at the hands of erroneous automated trading code, Knight Capital is fighting for its survival in what’s looking like a cruel twist of fate meted out by the automated market. With its future uncertain, Knight reportedly has opened its books to potential acquirers, including bulge-bracket firms.
Before this week, Knight was regarded as a powerhouse in electronic market making and institutional brokerage. But the firm’s so-called "Knightmare" started almost as soon as the markets opened Aug. 1. As trading unfolded, a faulty algorithm fired off millions of shares of erroneous orders and created havoc in 150 stocks routed to the New York Stock Exchange, rattling already weak investor confidence in the system. Early that morning, the company said it was having a “technical issue” on its listed market making desk that was affecting orders routed to the NYSE, adding that it had notified clients to route orders through other brokers. Even though Knight unwound the positions, and said no clients were harmed, some major retail customers — E*Trade, TD Ameritrade, Vanguard and Fidelity — routed their orders away from Knight, at least temporarily, curtailing revenue when Knight needed it most and complicating any potential acquisition of the firm.
According to today’s New York Times Knight rushed out a new system to participate in the NYSE’s new Retail Liquidity Provider program, without adequately testing it, though rival firms waited to take their time. The incident has raised questions about the electronic broker’s procedures for testing software and quality control and why it didn’t have a “a kill switch” to stop the out-of-control algorithm.
Live By the Sword, Die By the Sword
How quickly circumstances can change the fortunes of a successful trading enterprise in today’s high-speed ecosystem. The frenetic system is so complex and interdependent — with 13 exchanges and 50 dark pools — that when these sudden disasters strike, humans seem unable to turn off the machines. Even though the NYSE traced the problem to Knight, it didn’t have authority to halt trading within the first 15 minutes of the opening. In a cruel form of justice, it was market participants — hedge funds, individual traders at proprietary trading firms and asset managers — with specialized algos that detected the abnormal prices and acted swiftly to punish Knight for its careless software error. In the absence of regulators to safeguard the market or mandatory standards to guide electronic trading, the market seemingly evened the score.
Since this is the third disaster involving electronic trading and flawed software in five months — it occured on the heels of BATS’ halting its IPO in March due to a system glitch and Nasdaq’s technical problems with the Facebook IPO in May — on top of the May 2010 Flash Crash, even analysts and pundits who have vigorously defended the market’s reliance on high-frequency trading have begun to worry. Even they are acknowledging that the markets are too complex and too fragile to withstand these jolts. Still, out of fear that it would reduce profits or drain liquidity, the industry has resisted any proposal that would rein in computerized trading.
Better real-time technology could help regulators, contends David Leinweber, author of "Nerds on Wall Street: Math, Machines and Wired Markets" and the head of the Lawrence Berkeley National Laboratory Computational Research Division’s Center for Innovative Financial Technology. Writing yesterday in Forbes.com, Leinwebber called for “centralized real-time monitoring of markets, like the FAA does with air traffic.”
Fortunately, the SEC already is moving in this direction. But it has a long way to go. Last week, the SEC announced that it had licensed a high-speed market data research platform from Tradeworx, a high-frequency trading firm, to help the regulator oversee the equity and options markets. It’s also working on the larger, consolidated audit trail system. In the meantime, organizations like FIX Protocol Limited, whose buy- and sell-side members understand algorithmic trading, recently upgraded its risk control guidelines for brokers to prevent an errant algorithm or fat-finger mistake from blowing up the market, as Knight’s error did. And the SEC’s limit-up, limit-down rule for stocks is supposed to go into effect later this year, to help exchanges control abnormal price swings.
But tougher measures may be needed. After the Knight debacle, one buy-side trader said there could be calls for more regulation of algorithms. Obviously, no one wants to submit their algos to a federal agency for testing, but safeguards clearly are needed. Otherwise, how can themarket win back investors, when even the professionals’ nerves are shot?