Internet Bots: Anatomy of a Stock Selling Frenzy

In case you missed the brouhaha on Monday, United Airlines' stock dropped from $12.50 to $3 per share in the matter of 15 minutes because of an Internet bot--and not just any old Internet bot--the king of all bots, the Google News crawler. The circumstances that led up the event were confusing, fingers were pointed every which way, and conspiracy theorists cried stock manipulation. Now that dust has settled somewhat, the pieces of the puzzle have started to come together in this odd series of events.

Apparently the origins of this butterfly effect date back to December 10, 2002, when the Chicago Tribune published a story titled, "UAL Files For Bankruptcy," about United Airlines' 2002 bankruptcy filing. The story made its way into the online searchable database of the Sun Sentinel, where it remained available ever since.



Credit: Google

Sometime between 1:00 a.m. and 1:36 a.m. EDT on Sunday, September 7, 2008, a link to this old article appeared under the "Popular Stories Business: Most Viewed" section of the Sun Sentinel's site's business section. This section is dynamically updated based on the traffic volume of stories being viewed. The Wall Street Journal reports that this old story resurfaced when a "single late-night" visitor viewed the story on the site during a "low-traffic period."

 
 Credit: Google
At 1:36 a.m. EDT, the Google crawler discovered the new link on the Sun Sentinel site and noted that the link was not there previously when the Googlebot had last crawled the site 19 minutes earlier. The crawler interpreted this as a new link and followed the link to the story. The Tribune states that "no new story was published and the old story was not re-published -- a link to the old story was merely provided." Unfortunately, there was no date on the story itself, so the Googlebot used the only date it could find on the page as the context for the article, which was the actual date, September 7, 2008, as it appeared in the header of the page for the Sun Sentinel site (see the image to the left). Google states that "the article was indexed and then available through Google News search, but was not shown on our headlines pages."

At some point between Sunday morning and Monday morning, the story had been picked up by the investment and advisory research firm, Income Securities Advisors Inc., and then passed on to the Bloomberg financial-news service. The Silicon Alley Insider reports that "a non-detail-oriented securities analyst summarized the article (apparently without reading it first) and uploaded the summary to Bloomberg."

 
 Credit: Google
The Wall Street Journal reports that at about 10:45 a.m. EDT on Monday, September 8, "a headline from the outdated report flashed across Bloomberg screens." At which point, "algorithmic trading mechanisms, which buy and sell stocks based on news headlines and earnings data," started selling off the United Airlines stock on the mistaken assumption that the company had just filed for bankruptcy. United's stock plummeted over the course of the next 15 minutes until NASDAQ halted trading of the stock and United issued a statement saying that the rumors of its recent bankruptcy were greatly exaggerated. The Wall Street Journal adds that "UAL's stock price ended Tuesday's session at $10.60, down 2.8% on the day and nearly 13% off Monday's open."

Let's recap... An automated system added an old link to a dynamic Web page. An automated system detected the link as a new story and cataloged it as a current event. Enter a human for the only part of this tale: an analyst picks up the story and passes it on, not realizing that it is old news. Automated trading systems start a massive stock self off based on the old headline...

Call us doomsayers, but perhaps we've just witnessed the birth of Skynet.
Tags:  Atom, Internet, Stock, Tom, NATO, NAT, bots, bot, K
Comments
shanewu 6 years ago

Two lessons here...

1. Stories that aren't dated suck! There are a lot of sites that don't put obvious dates on stories and that seems asinine to me.

2. People trust google too much. :)

Post a Comment
or Register to comment