When software kills

This week, a software glitch essentially killed a company that had become a major force on Wall Street in the last 20 years. Knight Capital is one of the largest trading organization on American stock exchanges and lost a staggering $440 million in a bit over a half hour on Wednesday morning, pushing the company to the brink and driving some of its largest customers away and the cause of it all was a set of software bugs.

High speed trading

For the last few years, there’s been a war in the stock exchange world, with each side trying to outsmart the other through software. In the financial world, the phenomenon is known as high speed trading, or high frequency trading, where stock purchase and sale orders are sent and matched in milliseconds. Along the way, each wall street firm has work to find ways to execute trades faster and in larger quantities to gain an advantage over their competitors.

In this world, engineers and mathematicians develop increasingly complex formulas to shave a few cents of profits on a transaction. It’s a world where software and hardware combine to deliver results that can be counted in millions and billions of dollars; a world where the finest algorithm is the best weapon; a world where co-locating a computer server a couple of feet closer to the exchange machines comes at a premium; a world that is not too far away from wall street physically but eons away from the world of trading that existed only a couple of decades ago.

In this world, speed of trade execution matters and in that world, few were as fast or executing as many trades as Knight Capital Group. On any given day, between 15% and 20% of all trades on the NYSE and NASDAQ went through Knight Capital’s computers.

In early July, the New York Stock Exchange (NYSE) received SEC approval for a new program called the Retail Liquidity Program (RLP). The program, which was slated to kick off on August 1st, would allow retail traders to get access to better pricing on stock. The specification for the programme had been around for less than a year but a lot of wall street firms had planned for its implementation. There was some controversy when the announcement was made but all market makers started preparing for the launch, realizing that this may be another tool to fight the competition and that those who were not ready by launch time could risk losing out precious advantages in the market.

Knight Capital, as one of the largest players in the market, decided that it would provide that capability to its customers as soon as it became available and that it would take advantage of the price differential for its own trading. With only a few weeks from the announcement to the release, internal development teams must have burned the midnight oil to get this in place by the August 1st deadline.

Knight goes down

But on Wednesday morning, Knight Capital may have made a mistake that could cost it its existence. According to reports, the company had just finished upgrading its software to the newer version, allowing it to take advantage of the RLP. When the NYSE opened, trades on as many as 114 companies started running at higher than expected volumes. Here’s how the New York Times describes what happened next:

A New York Times analysis of New York Stock Exchange volume on Wednesday morning showed that during the first minute of trading there was 12 percent more trading in all stocks than there had been on average during the previous seven days. By the third minute of trading there was 116 percent more trading than the previous week’s average. The difference reached a peak at 9:58 a.m., when the volume was six times greater. After that, trading volume fell off sharply, nearing the recent average at 10:15 a.m.

Mr. Niederauer said that the exchange had noticed the problem and contacted Knight “within minutes” of the 9:30 opening bell.

Knight’s failure to respond sooner was particularly mystifying to other traders because on Wednesday the firm had introduced new trading software. Industry experts said that this would normally be cause for programmers and other employees to be on high alert.

Once the problems began, many traders said it would have made sense if the firm’s employees had not caught the problems for the first minute or so, given the speed at which Knight’s program was firing off orders. After that, though, the problems were visible for all to see.

In under an hour, the company had lost $440 million, $75 million more than all the cash it had in its reserves, or more than it had made in profit for the previous year. Customers, wary of suffering from the bug, decided to start trading through other firms. By Thursday, a number of brokerages and broker-dealers (TD Ameritrade, Scott Trade, Fidelity Investments, Vanguard Group, E*Trade and Pershing LLC, a division of BNY Mellon) suspended routing trading orders through the company. A few came back but not all.

Meanwhile, investors in the company sent its price tumbling down, along with the rest of the stock markets’ as confidence in the well oiled machinery of stock trading was shaken. The trading losses forced the company to open itself up to the prospect of either bankruptcy or a sale, with many potential suitors circling.

For all intents and purposes, Knight will come out of this as a significantly smaller company, if it remains a standalone company at all. Meanwhile, it has become a cautionary tale to the rest of the industry, leaving many in the financial world wondering if they could be next.

A dangerous pattern

The Knight Capital failure, however, is not the first one related to software in the financial world. A couple of years ago, software was seen as the power behind a flash crash when the Dow Jones lost nearly 10% of its overall value in a few minutes. Last March, the dollar lost 5% of its value against the yen and the price of Cocoa futures lost almost 15% of their value, both incidents that were results of high frequency trading.

Meanwhile, BATS, one of the secondary markets, tried to go public in late March but software Snafus force the company to cancel its IPO and even impacted the price for Apple’s stock. And even the most anticipated IPO of the year, Facebook, was marred by technical glitches as the NASDAQ.

With 60% to 70% of all stock trades being executed by automated computers, there is growing concern that software is increasingly threatening the stability of our markets. Along the way, volatility has gone through the roof, which can send algorithms on large sale cycles that can bring markets increasingly down.

The trend will continue to growing automation and there is a widespread feeling around the financial world that this is just the beginning of these types of phenomenon. In 2008, I started wondering if volatility was the new normal; in 2012, I am increasingly convinced (and concerned) that algorithms will not only lead to increased volatility but also down a path where few people may be understanding how financial markets are functioning, potentially detaching them from the realities of economics fundamentals.

Because so many people, from the corporate world to the political one and everyone in between, are looking to stock markets as indicators of the health and wealth of economies, we may find ourselves in a situation where the economic indicators are so divorced from any fundamentals to software could end up not only killing companies (as it is doing for Knight Capital) but possibly even take nations down with them. And that is the dark side of dark pools.

About the Author

Tristan Louis

Writing and working on the internet since 1993, I've launched six companies, of which two went public and three were sold. This is my personal site and all opinions here are mine.

  • Steven Baker

    These examples don’t kill people… at least directly. I was expecting software bugs in flight control systems etc

  • http://www.tnl.net Tristan Louis

    … but they do kill companies, don’t say? I didn’t say that software killed people :)

  • http://www.spinxwebdesign.com/ Spinx Inc

    The software for financial industry is risky and its failure really kills many interrelated systems.