Market maker Knight Capital Group announced this week that it is taking steps to improve its operational risk management after an electronic trading glitch in August cost it $440 million in about 45 minutes. It initially blamed “old dormant software.” The losses forced the firm sell more than 70 percent of its preferred stock for $400 million in order to stay in business.
The Wall Street Journal reported that Knight’s CEO Thomas Joyce will be appointing a chief risk officer to oversee both market credit and operational risk issues. In addition, Joyce said that the firm hired IBM in late August to investigate its software development practices and processes and report its findings to the board later this autumn.
The WSJ article was more cloudy than clear when it came to the cause of the glitch. Joyce was reported as saying that the trading glitch was caused by an “undetected bug” in old software that allowed other software to generate trading orders unrestricted by volume caps. The bug was apparently triggered when Knight’s new electronic trading software was installed improperly.
To say the least, this explanation is a bit bizarre. A latent bug in what once was operational software was triggered by a poor installation of new operational software? IBM will earn its money if its report explains how that happened.
Joyce, saying he was “deeply embarrassed” by the episode, then blamed it all on the company’s IT group. According to the Journal, Joyce said, “People do stupid things… A small team of people made a grievous mistake.” Joyce later went on to call them “knuckleheads.”
Maybe before top management blames internal IT "knuckleheads,' they should be looking inward. I wonder how much pressure Joyce—who has been an “unapologetic advocate” of automated trading—and his management team placed on the company's IT group to hurry up and get that new trading software installed. I also wonder how open Joyce would have been to his IT group telling him that they needed more testing time. I suspect not much.
The SEC has tried to keep rogue algorithms from creating havoc by mandating “circuit breakers” to halt sudden and inexplicable spikes in a stock’s price, but as this case showed, that isn't sufficient. In an attempt to forestall more regulatory oversight, the stock exchanges and the market markers are now looking at the possibility of installing “speed bumps” that “would stop all orders from one market maker at an exchange if the situation called for it, and also possibly across multiple exchanges,” a Money News article yesterday reported. In addition, they are discussing the creation of a “kill switch” to “shut down order flow” when a trading algorithm begins to go rogue.
Joyce is not enamored with market kill switches, however, since they might shut down all of a firm’s trading when only one trading area is being affected, he says. I find this an interesting position to take as there was no effective kill switch in the company’s own software to turn off the errant trading that almost sunk his firm. If there had been a market kill switch, he might have still been "deeply embarrassed" but the firm also may have lost a lot less money.
Joyce went on to say that he expects the debacle to eventually make the company smarter and stronger. Given that Knight Capital’s stock closed yesterday at $2.62, the lowest it has been since the bailout (and down from $10.33 before the glitch), investors seem to think it won't be soon.
Contributing Editor Robert N. Charette is an acknowledged international authority on information technology and systems risk management. A self-described “risk ecologist,” he is interested in the intersections of business, political, technological, and societal risks. Along with being editor for IEEE Spectrum’s Risk Factor blog, Charette is an award-winning author of multiple books and numerous articles on the subjects of risk management, project and program management, innovation, and entrepreneurship. A Life Senior Member of the IEEE, Charette was a recipient of the IEEE Computer Society’s Golden Core Award in 2008.