Glitch Google News caused a collapse in stocks: how it happened
An amazing case occurred on the US stock market on Monday. In the financial newsletter Bloomberg got the news that the airline United Airlines filed for bankruptcy. The bear market reacted predictably and the stock price instantly collapsed. Within a few minutes, the total capitalization of the company fell by about $ 1 billion. Only after five to ten minutes, people figured out what was wrong: it came to the traders that this news was actually six years old, which got into the mailing list by mistake. The course, of course, quickly recovered back. If you know about such a movement of quotations in advance, then you can make huge money on such a sharp movement.
How could news from a forgotten archive get into the hot news feed? This would not have been possible without a series of fatal coincidences, errors in the algorithms of online services and human errors. If we assume that someone planned the entire chain of "coincidences" in all details, then this is simply a genius scam that brought him tens of millions of dollars in a few minutes. And to prove his guilt here will be absolutely impossible.
This story also sheds light on the mechanisms for disseminating information on the Internet and on the potential vulnerabilities of the online media space, as well as fully automated trading on the stock exchanges.
That's how it all happened .
')
The text of the press release with a statement about bankruptcy, as it turned out, is the content of the Chicago Tribune newspaper for December 2002.
It all started with the god of the forgotten site of the small Florida rag Sun Sentinel, which belongs to the same media holding as the Chicago Tribune. Since this is a common holding, the Chicago Tribune archives are available from the Sun Sentinel website. Articles are physically stored on their server and have been in the public domain for six years. But they lie in that part of the site, where it is forbidden to enter search bots.
However, on Sunday, the text of the mentioned press release was included in the list of the most popular articles on the Sun Sentinel website (it is unknown why this happened, but the site’s attendance is small, so there’s nothing surprising in this random surge of popularity). Thus, the link to the archived material went to the Sun Sentinel home page, which is indexed in Google News.
And here was the first glitch. Algorithms of Google News considered that the material of 2002 is a “new material” and placed it in the section of new materials in its aggregator news.google.com. Obviously, the Google News system did not take into account the date of the news, but simply in the search index of this text was not, and therefore the news was considered "new."
Naturally, Google denies the charges. They explain that on the Google News site the news was in the “new” section, because it is really fresh material that was not there before. In this case, the text of the press release was clearly date: 2002. If one of the people did not pay attention to it, then this, they say, is already their problem.
What happened next is a human error. The editor of Bloomberg saw the “fresh” news and included it in the hot news feed that is distributed among traders. Naturally, in the trading business, seconds are taken into account, everything is done in a great hurry. You can understand the editor, who weakened the vigilance and trusted authority of Google News.
We also add that such a sharp drop in stocks would be impossible if the traders were people. But nowadays more than half of transactions on the stock exchange are carried out automatically - special trading programs that instantly catch the slightest movement of quotes and activate the programmed action algorithms. As soon as the fall in quotations began, in these programs, triggers to dump stocks began to work instantly. A wave of asset discharges escalated like a snowball. It all happened in a few minutes.
How could news from a forgotten archive get into the hot news feed? This would not have been possible without a series of fatal coincidences, errors in the algorithms of online services and human errors. If we assume that someone planned the entire chain of "coincidences" in all details, then this is simply a genius scam that brought him tens of millions of dollars in a few minutes. And to prove his guilt here will be absolutely impossible.
This story also sheds light on the mechanisms for disseminating information on the Internet and on the potential vulnerabilities of the online media space, as well as fully automated trading on the stock exchanges.
That's how it all happened .
')
The text of the press release with a statement about bankruptcy, as it turned out, is the content of the Chicago Tribune newspaper for December 2002.
It all started with the god of the forgotten site of the small Florida rag Sun Sentinel, which belongs to the same media holding as the Chicago Tribune. Since this is a common holding, the Chicago Tribune archives are available from the Sun Sentinel website. Articles are physically stored on their server and have been in the public domain for six years. But they lie in that part of the site, where it is forbidden to enter search bots.
However, on Sunday, the text of the mentioned press release was included in the list of the most popular articles on the Sun Sentinel website (it is unknown why this happened, but the site’s attendance is small, so there’s nothing surprising in this random surge of popularity). Thus, the link to the archived material went to the Sun Sentinel home page, which is indexed in Google News.
And here was the first glitch. Algorithms of Google News considered that the material of 2002 is a “new material” and placed it in the section of new materials in its aggregator news.google.com. Obviously, the Google News system did not take into account the date of the news, but simply in the search index of this text was not, and therefore the news was considered "new."
Naturally, Google denies the charges. They explain that on the Google News site the news was in the “new” section, because it is really fresh material that was not there before. In this case, the text of the press release was clearly date: 2002. If one of the people did not pay attention to it, then this, they say, is already their problem.
What happened next is a human error. The editor of Bloomberg saw the “fresh” news and included it in the hot news feed that is distributed among traders. Naturally, in the trading business, seconds are taken into account, everything is done in a great hurry. You can understand the editor, who weakened the vigilance and trusted authority of Google News.
We also add that such a sharp drop in stocks would be impossible if the traders were people. But nowadays more than half of transactions on the stock exchange are carried out automatically - special trading programs that instantly catch the slightest movement of quotes and activate the programmed action algorithms. As soon as the fall in quotations began, in these programs, triggers to dump stocks began to work instantly. A wave of asset discharges escalated like a snowball. It all happened in a few minutes.
Source: https://habr.com/ru/post/39565/
All Articles