Automated News Crawling Evaporates $1.14B 546
cmd writes "The Wall Street Journal reports that Google News crawled an obscure reprint of an article from 2002 when United Airlines was on the brink of bankruptcy. United Airlines has since recovered but due to a missing dateline, Google News ran the story as today's news. The story was then picked up by other news aggregators and eventually headlined as a news flash on Bloomberg. This triggered automated trading programs to dump UAL, cratering the stock from $12 to $3 and evaporating 1.14 billion dollars (nearly United's total market cap today) in shareholder wealth. The stock recovered within the day to $10 and is now trading at $9.62, a market cap of $300M less than before Google ran the story." The article makes clear that Google's news bot only noticed the old story because it has been voted up in popularity on the site of the South Florida Sun-Sentinel newspaper. The original thought was that stock manipulation may have been behind the incident, but this suspicion seems to be fading.
What if Google evaporates itself? (Score:3, Interesting)
It would be so awesome if the Google news bot was able to dig up some old dirt on Google that would make it's ~$450 per share stock crater... that would be an awesome site to behold.
I doubt it'll happen, since there haven't been any previously bad financial reports that could come back and masquerade as current news, but you never know.
It also serves to show how fragile our financial system is (just read the "what if" reports on what would have happened if Fannie Mac/Freddie Mae were allowed to go under... it would have been BAD). Sure, United gained back most of its value but it's still down a good bit of cash. It reminds me of when Abbie Hoffman threw a bunch of $1 bills onto the NYSE trading floor and TRADING STOPPED as the floor traders ran around picking up the bills.
BBC News - Most Read (Score:2, Interesting)
I've noticed a few time lately that the BBC News site's 'most read' list contains stories from years ago. It's very misleading because the links are on the main news page and the date, although present at the top of each story, is fairly subdued. I've been caught out a couple of times by this. Of course, once a story gets on the 'most read' list everyone clicks on it pushing it further up the list and prolonging its being there.
Hello? Is this thing on? (Score:5, Interesting)
Google News crawled an obscure reprint...
The story was then picked up by other news aggregators...
This triggered automated trading programs...
Is there even a live person at the wheel anymore? Or is SkyNet just fucking with us now?
Re:Holy crap. (Score:5, Interesting)
This does show just how fragile a system can be when there is a strong disincentive to going second or third on tasks that one would normally think you should have human interaction.
Re:Holy crap. (Score:5, Interesting)
The next Ponzi scheme? (Score:3, Interesting)
That said, are automated trading systems going to be this decade's great Ponzi scheme? I can't believe so many people can be so lazy with their investments to send the stock tumbling that much.
It looks like the ones who'll ultimately get burned from this are are those who are careless with their money. But how soon before people take advantage of viral networking to manipulate Google's algorithms for determining popular news, bring up old doom and gloom articles to intentionally tank a stock so they can buy while it's cheap?
Still sounds like manipulation to me (Score:5, Interesting)
So are you telling me that I could set up a "click-bot" to vote up old-news and make myself rich in the ensuing mayhem?
Trading pattern is striking (Score:4, Interesting)
Wow. Take a look at Monday's trading history for UAUA. [yahoo.com] Look at that drop. And notice that it happened on huge volume; several hundred million dollars changed hands within fifteen minutes. It wasn't just a few traders running the price down in light trading.
The stock hasn't come back all the way. It's still down 20% for the week.
Here's the newspaper page that started it all [blogspot.com], as archived by Google.
Missing Dateline (Score:5, Interesting)
Another point that no one seems to be picking up on is the problem of a lot of news sites neglecting to include a dateline on their stories. I've run in to this a number of times, and it makes it difficult to determine the relevance of a given story. It's a very simple thing to include the date the story was published, but a lot of sites don't seem to bother.
Re:Holy crap. (Score:4, Interesting)
Yep only if these stupid people are not these who manage your pension fund ? Because they are of these who use this kind of software...
Re:Holy crap. (Score:3, Interesting)
Joke Economy (Score:5, Interesting)
When a blip at a local Florida newspaper can combine with a trivial bug to destroy a major American airline in a morning, the economy and the "reporting" that the economy depends on is revealed to be a giant joke.
This episode was remarkable because it was huge, fast, and concentrated in a single high profile corporation. But how much of this broken system's smaller problems go "unnoticed"? Unnoticed as problems, at least, but showing up in all kinds of market valuations and economic decisions based on them that are all built on a landscape of errors, omissions and misunderestimations?
How can you trust an economy that makes mistakes like that? Anyone smart would find any alternative that's less crazy and put their money into it.
Re:Holy crap. (Score:4, Interesting)
That's not how the systems work.
It's PEOPLE who decide to dump stock based on the headline. Dumping stock makes the price go down. The SOFTWARE part of the story is called a "stop-loss order", which is an automatic "sell" order based on stock price. The theory, which works in practice, is that dumping a stock that has lost 50% of its value (or some other trigger level) will prevent a complete loss of money on that stock. Half of what you had is better than 10% of what you had.
It replaces the need for someone to sit at the computer monitoring the stock prices every minute of the day. It also prevents the time delay (and loss) for someone who does monitor prices every minute having to enter and execute the sell order.
If they need something, perhaps going after those companies for artificially deflating the stock's value would be the best course. It's not like this couldn't have happened with humans.
That's right -- the source of the story should be liable. And it's not like this couldn't have happened with humans, because it started with humans.
Now, as someone else commented, it's day trading (or short term) that causes this kind of thing. If you buy stock and sit on it (not literally), it will usually go up. My $5 Sony stock has split once and is much higher. Of course, I bought it 30 years ago. And only ten shares. Sigh. It's been higher than it is now, but it's still higher than when I bought it.
This kind of information tempts one to say "make a law" that all stock bought must be held for six months before it can be sold. That's still not a solution. It will just move the problem over to options, and keep short term money out of the market, making it smaller.
Re:BEHOLD.... (Score:2, Interesting)
What would make this story more sinister would be if it turns out that someone at Google bought a bunch of the stock while the prices were low. Now *that* would be interesting.
Re:Holy crap. (Score:3, Interesting)
Note that it went onto Bloomberg (which is what caused all the traders to freak) because a human put it there. Without fact checking.
Re:Holy crap. (Score:3, Interesting)
I'm all for letting stupid people lose their money. What I am against is letting stupid people lose other people's money. especially over something as idiotic as this.
Anyone who lost a significant portion of someone elses money over this should be banned from trading with other people's money. I'm sure more then a few brokerage firms lost clients money.
Re:Holy crap. (Score:1, Interesting)
First off, I've never heard of a trading bot "read the news" - it wouldn't be profitable to dump every stock mentioned in an article with the word "bankruptcy." Most business news articles will talk about many companies at a time, and it's just dumb.
What happens is when a human sells an abnormal amount of stock, then the bots notice something weird is going on, through volume or price changes, and the bots are typically more of a fail-safe mechanism. If they see some stock plummeting and they aren't told that it's ok, they close out any exposure to it to reduce risk - who cares why the knife is falling, just get away from it.
On the other side, some bots are programmed to take risk trading momentum - if a stock starts climbing, then the bot will try to get in on the action. Or if a stock is falling, it might try to short it. These bots are less common because you have to supervise them, as errors would ruin many careers.
The fact that it was programmed trading that brought the stock down gives a high probability that the stock will go up tomorrow. Stop-loss bots heavily outweigh risk taking bots, are less quickly supervised, and any bots not told to ignore yesterday's data will think the risk level on UAL is dramatically higher than it really is. Of course, the human traders also know this, and you the lay person have effectively zero chance beating them to the profits.