Sunday, November 11, 2012

Spotlight: Knight Capital

Two weeks ago, I covered the Flash Crash of 2010. What was significant about the Flash Crash was the way the market had fallen so far in a span of only minutes, but we have to remember that the market recovered almost just as quickly. Individual firms and stocks only suffered net losses of a few percent at most. Additionally, the Flash Crash impacted lots of HFTs and so the losses were shared by the market as a whole.

What would happen in the possible scenario where some computer glitch could corrupt an individual trading firm's or hedge fund's HFT algorithm?

Fast forward to August of this year.

Knight Capital Group, a trading firm founded in 1995, had over the past few years grown to be one of the biggest "middle-men" in the market. Knight executes trade orders for retail brokers like TD Ameritrade and ETrade, and competes with the likes of UBS, Citibank and Citadel. It is headquartered in Jersey City, NJ, and had employed over 1,400 people in 21 locations around the world.

Up until this catastrophic event, according to Adam Sussman at the data company Tabb Group, Knight Capital was responsible for 11% of all trading in American stocks between January and May. This is no small fish!

On the morning of Wednesday, August 1st, crisis struck. Somehow, Knight's automated stock program flooded the market with trades. In the first 45 minutes that the market was open, a computer glitch in their algorithm caused the firm to buy and sell millions of shares of hundreds of stocks, causing the value of these stocks to soar upward. Knight blamed this on their new trading software, which was installed that same week.

Since Knight was now holding onto so many shares of these stocks, they had to rapidly sell everything to minimize losses. The problem was, however, that their shares were overvalued because of their glitch. They had to sell their losing positions back to the market and take losses, which totalled $461.1 million.

According to the New York Times, that's $10 million lost per minute. Wow.

These total losses were greater than the revenue that Knight Capital earned in the second quarter of this year, which was $289 million. Because of this disaster, Knight had to seek help from a group of customers and rivals. The Wall Street Journal reported that it's share price was down 78% as of mid-October.

Thankfully, the rest of the market wasn't impacted. This could be due to volatility controls put in place by the SEC in the aftermath of the Flash Crash.

Some questions linger after the incident. Why didn't traders from Knight Capital step in and stop the rogue algorithm? Are the current safety regulations set in place by the SEC enough to stop glitches like this from happening again?

In my opinion, I think there's a very low chance of something like the Knight Capital incident happening again. The Flash Crash happened once, and now regulators have learned to put safety nets in place to save the market in such a scenario. We learn from our mistakes, so every other firm in the market will be sure to check for bugs in their algorithms once more because they don't want to be the next Knight Capital, losing millions of dollars.

Sources:

[1] Jacob Bunge, "Knight Plans for Worst and Hopes for Best", Wall Street Journal, October 17, 2012, http://online.wsj.com/article/SB10000872396390444734804578062174162145686.html

[2] Danielle Douglass, "Computer Glitch Sent Markets into a Tizzy", Washington Post, August 1, 2012, http://www.washingtonpost.com/business/economy/computer-glitch-sent-markets-into-a-tizzy/2012/08/01/gJQAwNsMQX_story.html

[3] "How Safe Is Computer Trading After Knight Capital Crash?", The Week UK, August 3, 2012, http://www.theweek.co.uk/us-business/48304/how-safe-computer-trading-after-knight-capital-crash#

[4] Nathaniel Popper, "Knight Capital Says Trading Glitch Cost It $440 Million", New York Times, August 2, 2012, http://dealbook.nytimes.com/2012/08/02/knight-capital-says-trading-mishap-cost-it-440-million/

1 comment:

  1. You seem to be very optimistic on the robustness of code. Many of these software projects are probably hundreds of thousands to millions of lines of codes, and the bugs that happen aren't the result of one individual piece of code -- they result from interaction between many different pieces of code. It's not just a matter of checking for bugs once more.

    How will another Knight Capital incident be prevented?

    ReplyDelete