High Frequency Trading: Markets, Practice and Models.

Only three years after May 2010, when 862,000,000,000$ were erased in market value, a new "Flash Crash" scared the United States and the financial markets around the world. On April 23rd, 2013, a false news about an explosion at the White House spread in few seconds in all the most important financial markets causing a cascade of selling. To have a better idea, you may think that in two minutes the Standard & Poor's 500 lost 136,000,000,000$.
Thanks to [Wang, Kisling, Lam. (2013): "Fake Post Erasing $136 Billion Shows Markets Need Humans"] we are able to explain what happened in those minutes.
At 1:07 p.m., the S&P500 was rising about 1% at 1,578 when the false news[2.12] began to spread after its publication on the Twitter account of "Associated Press".
The, already known VIX, at 1.08 p.m. had gained about 9.34% (from 13.6 to 14.87) in only two minutes, but when it reached its maximum at 1.10 p.m., it had lost all the gain in the following three minutes.
At 1.10 p.m., S&P500 stopped rising immediately and it reversed its trend, falling down to 1,563.
Fortunately, this time, the crisis lasted only three minutes. In fact three minutes after the news diffusion, Associated Press denied the previous (false) report stating further that its Twitter account has been violated by anonymous hackers and that, obviously, there hadn't been any explosions.
Thus, the index recovered from the plunge within three minutes, so that the S&P 500 closed its trading day with +1% at 1,578.78, while the Chicago Board Options Exchange Volatility Index losses of about 6.3% were registered, closing its trading at 13.48.
According to many traders it is very dangerous that software controls billions of dollars since the sells were started when algorithmic trading programs read news headlines. The most disturbing aspect of the second Flash Crash in history is that while the software believed the false news, the human traders immediately verified the report's authenticity.
In other words there is a reasonable suspicion that the selloff had been exacerbated by "stop-loss orders".
And as we have already said in the previous chapter, they are placed by HFTr in order to automatically sell stocks when a specified threshold (i.e. losses) is reached. [...]