Computer Decision Making Costs Knight Capital Group $440 Million

This last Thursday Knight Capital Group, Inc. disclosed that due to a computer “glitch”, a technology “snafu”, the company lost $440 million. Their shares sank 63% and closed at $2.58. And it took just 45 minutes. $440 million was almost twice what the revenue for the second quarter. The immediate question is, how can a computer glitch nearly ruin a company? Don’t they believe in stop losses?

What was the computer glitch that Knight Capital faced? Their algorithmic trading software sent erroneous orders to NYSE-listed securities, rapidly buying literally millions of shares in more than one hundred stocks, pushing the value of those stocks higher. The company was then forced to sell the higher-valued shares back to the market at the lower price, causing the extreme loss.

What is algorithmic trading? Algorithmic trading, otherwise known as robotic trading, or trading from a black-box, is the use of computerized decision making for initiating trades to the exchanges. The software, itself, decides on the security price, liquidity available, timing (how quickly the order can be filled), overall risks and costs associated with the trade. Computerized decision making, in the vast majority of cases, is done without human intervention.

Let’s put algorithmic trading into perspective to understand just how Knight Capital could lose $440 million in 45 minutes.

  • 1 second is 1,000 milliseconds
  • 1 millisecond is 1 million nanoseconds
  • One blink of an eye takes 300 milliseconds
  • The default for one mouse click is 500 milliseconds
  • The average human reaction time is roughly 250 milliseconds
  • For a professional chess player to decide on the next move is 650 milliseconds
  • From the moment that a ball leaves his opponents racket until it hits his own, a pro tennis player needs 500 milliseconds

In the time it takes to blink an eye, click a mouse, make a decision, or react, computerized decision making algorithms have received the datafeed from the exchange, analyzed it using various technical indicators, initiated their orders and been filled. If it takes less than 1 second to initiate trades and be filled, there is no possible way human oversight can stop “computer glitches.”

Exchanges have no one to blame but themselves. There are approximately 20,000 trading firms operating throughout the various exchanges today. Algorithmic trading firms comprise just 2% of all the firms. Yet these 2% represent 73% of the total daily U.S. trades. In a PBS documentary on March 15, 2012, Paul Solman reported that across the exchanges (NYSE, NASDAQ, AMEX, CME, NYMEX), the average trade time is now 22 seconds. A $300 million transatlantic cable is currently being constructed between New York City and London just to shave off 0.006 seconds. Physicists are now working on execution times at the speed of light, down to picoseconds (1 trillionth of a second).

Remember the Flash crash of May 6th, 2010 where a $4.1 billion automated sale triggered thousands of sells throughout the Markets. From that crash a study was conducted, led by Neil Johnson, a complex systems specialist at the University of Miami, and simulation engineer Brian Tivnan of the University of Vermont. They analyzed millisecond-scale price logs from 600 markets, using numbers gathered by Nanex live market data. What they discovered was that from 2006 to 2011, there have been roughly 18,520 sub-950-millisecond crahses and spikes, about 10/day. The Flash crash of May 6th was just waiting to happen.

Human decision making takes into account diverse aspects of the Market’s behavior, both technical indicators as well as fundamental interactions (overall market performance, whether it is an up or down day, emotional reactions due to news releases, etc.). What Tivnan noted was that algorithmic trading tends to “sacrifice diversity for speed”. He said, ” You see a lot more homogeneity at the sub-second scale than we see above 1,500 ms.” So many algorithms are likely to concentrate on a small set of strategies, having optimized for speed, resulting in algorithmic trading becoming vulnerable to systemwide “herd mentality”.

Is it any wonder then that a computer glitch could take just 45 minutes and cost a company $440 million?

Barbara Cohen CIO, Shadowtraders, and professional day trader, specializes in teaching students how they can be trading futures with their own trading system and trading strategies. Ms. Cohen has helped hundreds of traders achieve their goals trading. Find out if trading futures is for you by attending one of Ms. Cohen’s Free Webinars. Check out my Futures Trading Articles. For more information, send an email to or call 866-617-2037 today.

Visit our Futuresblog Partners:
  • Digg
  • Netvouz
  • description
  • ThisNext
  • MisterWong
  • Wists
  • Blogosphere News
  • Technorati

Tags: , ,

Leave a Reply

Security Code: