Join 3,512 readers in helping fund MetaFilter (Hide)


He didn't call the company, but he did write up a headline.
September 10, 2008 4:01 PM   Subscribe

"News-flow" analysis is used by some automated trading systems to buy and sell stock based on stories circulating in the media. United Airlines (UAUA) stock plummeted the other day. And it plummeted fast. Like, lost %75 of it's value in 20 minutes fast. The problem? The news-flow that triggered this sell off was based on a mis-labeled story from 2002. via /.
posted by butterstick (18 comments total) 6 users marked this as a favorite

 
Pepsi pepsi pepsi pepsi pepsi pepsi. Pepsi pepsi pepsi pepsi BAD pepsi pepsi.
posted by spiderwire at 4:29 PM on September 10, 2008


i'm not sure it's a good post, but i did think it was an interesting news story. and i don't see how it is pepsi blue.
posted by snofoam at 4:36 PM on September 10, 2008


I find this topic very interesting and I don't consider this Pepsi Blue, however, I seriously doubt this was 100% due to algorithmic trading.

There are tens of thousands of people that sit in front of their Bloomberg and Reuters terminals all day watching news live as it streams in. Once new news comes in from many wire services, a line will flash on their screen and traders will eyeball it and trade accordingly as fast as they can humanly can.

The goal of algorithmic trading systems based off news is to do the above faster and more accurately than people ever could.

My problem with this is that from what I've done in this space, it is highly unlikely that there's enough automated parties willing to take enough risk to cause a spike that big. The confidence numbers simply aren't big enough for these algorithms to be THAT confident. If all the aggregate algorithms out there bets money big enough to cause UAUA to go down that far, then they will lose money VERY quickly in other trades, because the signal/noise ratio is way too big to make huge bets off ONE new item (no matter how big the confidence number is on that single item). There's no way it can be that confident statistically. I could see it if they had about 10 articles and the wires picked up each others story over the span of several hours, but that's not what happened. The automated algorithms I see that WOULD do this are rules/technical based trading (which doesn't look at news at all, only price action), e.g. "SELL if it drops more than 10% in the span of 10 minutes), which would exacerbate the problem of a bunch of traders manually dumping the stock.

I'd be VERY surprised if it wasn't just a bunch of traders panicking in front of their Bloombergs and calling each other freaking out over the span of minutes.
posted by amuseDetachment at 4:45 PM on September 10, 2008 [1 favorite]


Yeah, what the fuck was that? On any other day, that probably would have been the top business story, but everyone was busy watching FNM and FRE collapse while the market soared. I feel sorry for all the investors that had their stop-loss orders executed, but man, when something hits those Bloomberg terminals, you really get to see the efficient market at work.

So it looks like the Sun-Sentinel committed the initial blunder by placing the old story on their website (although they claim it was in an "archive" section). Income Securities Advisor took the undated article and published it to Bloomberg without any fact-checking or verification, and at that point it's game over, the entire market thinks UAUA is filing Chapter 11. I don't think Income Securities Advisor has much of a future providing services to Bloomberg, and I can smell the class action all the way from here.
posted by malocchio at 4:54 PM on September 10, 2008


Fascinating post. Thanks for this. I was totally unaware of algorithmic trading.
posted by willie11 at 4:54 PM on September 10, 2008


I think of the stock market as a massive, expensive mosh pit.
posted by RobotVoodooPower at 5:31 PM on September 10, 2008 [1 favorite]


and I can smell the class action all the way from here.

When they discussed this on NPR yesterday the legal expert said there there is almost no way to recover money from decisions based on bad reports in news articles. News orgs make mistakes, we all know that. You are responsible for judging information for yourself.
posted by a robot made out of meat at 5:42 PM on September 10, 2008 [1 favorite]


I don't think spiderwire is saying this post is Pepsi Blue, rather making a joke that the "news" is vulnerable to malicious pollution as well, with a 'bad' in the midst of the Pepsi.
A bit like the geeks who try to break Echelon
posted by bystander at 6:44 PM on September 10, 2008


Holy freaking shit!! I still remember the time a Sony rep and I managed to miscommunicate somehow (I still maintain he said what I reported, he claimed I misunderstood him) and I managed to drive Qualcomm's stock down by a few points until they got a correction out on the wires.

(This was back when they were partnering to make CDMA cellphones and I reported that Sony was making some new ones on its own without QCOM. This was interpreted as Sony deciding it didn't need QCOM's help anymore and backing away from the alliance which would have hurt Qualcomm's market access if true.)

It gave me cold sweats. I can't imagine having literally destroyed 75% of the value of UAL in 20 minutes by miscoding a database entry. Holy shit...
posted by Naberius at 6:47 PM on September 10, 2008


The stock cratered to three dollars in 4 minutes, then recovered 75% of value in the next 4 minutes. If they were human trades, that's some fast phone-calling.
posted by anthill at 8:12 PM on September 10, 2008


Not all phone-calling, some people can trade online.

Like me.

Oy!
posted by mazola at 10:31 PM on September 10, 2008


The Wall Street Technology article is interesting but as amuseDetachment points out there are many other reasons why UAL plummeted so far so fast.

Driving trades with XML type of feeds is just the newest enhancement to Program Trading, a practice that's been in place on Wall Street since the late 80's.

In terms of the magnitude of the drop, although I haven't read definitive explanation yet I suspect it had far more to do with Stat Arb or Volatility Arbitrage than with this news driven trading.

Statistical Arbitrage is interesting and if you get it right, a steady - and at times large! - money maker. But most stat arb techniques focus on shares, identifying pricing mismatches between similar stocks and exploiting the same. While we have observed broad sell offs in the market that later were found to be caused by Stat Arb, I don't think this was an example as it went too far, too fast.

Lately we've seen lots of attention focused on Volatility Arbitrage, trading across asset classes (e.g., shares and derivatives such as options). A brief explanation.

The price of a stock option is determined by several variables, one of which is the volatility of the underlying shares. By holding other inputs to the pricing formula constant (e.g., interest rates), we can determine what volatility is priced into the option. This is known as Implied Volatility.

If we observe a significant difference between volatility, as embedded in the price of the option (i.e., "implied volatility"), and the volatility as directly observed in the stock market, we can structure trades to exploit this difference. This is, by definition, Volatility Arbitrage, arbitrage, of course, referring to the simultaneous sale or purchase of the same asset in different markets, to exploit pricing differences. In this case, volatility itself was the asset being bought or sold.

We know that "Many fund managers have been strengthening their capabilities to trade this previously ignored asset class.", the quote in question referred explicitly to volatility.

So I'd bet that most of this action was caused by automated trading, but focused on volatility plays. Sure, the press release kicked off all the fun, but traders were really focused on Volatility, not the news.

You say you didn't like this ride? Well, get ready for more, as some of the big funds are now targeting volatility.

Interesting and timely finance FPP - many thanks!

Whether or not Volatility Arbitrage was indeed the culprit, if you're interested in the topic here are some additional resources I had about:

A Volatility Arbitrage [.pdf] factsheet by Standard and Poors
A July 2008 paper by Standard and Poors [.pdf]
An interesting article [.pdf] from Hedge Fund World, 2003.

posted by Mutant at 2:56 AM on September 11, 2008 [4 favorites]


Nice post Mutant.
posted by JohnnyGunn at 3:41 AM on September 11, 2008


The more ways there are to take advantage of the market, the more ways there are for people to manipulate the market. In this case it was inadvertent.

Making trades faster doesn't mean making better decisions. Just means more volume, which means more money only for the traders.

Imagine if there were a "Litigation Generation Algorithm" that scanned news headlines and generated lawsuits and automatically filed them. Then you'd have to have a whole battalion of lawyers just to fix the damn errors.
posted by Xoebe at 6:32 AM on September 11, 2008


When they discussed this on NPR yesterday the legal expert said there there is almost no way to recover money from decisions based on bad reports in news articles. News orgs make mistakes, we all know that. You are responsible for judging information for yourself.

Yeah, after digging around a bit, I'll happily recant that part of my rambling. Here is a ruling that addresses most of the issues, and the shareholders don't appear to have much recourse.
posted by malocchio at 3:35 PM on September 11, 2008


You say you didn't like this ride? Well, get ready for more, as some of the big funds are now targeting volatility.

Pilot induced oscillation for the win!

While using such techniques on the stock markets is old news, and was inevitable from the start, it is also terrifying.
I did my Master's in Control Systems
posted by Chuckles at 7:53 PM on September 11, 2008 [1 favorite]


I just took a closer look at that wiki page.. There is something very wrong with the person who devised a linear scale to rate the effects of overly optimistic linearization. I mean, "pilot input causes divergent oscillation" is only 3x worse than undesirable motions which do not compromise task performance, right?
posted by Chuckles at 8:59 PM on September 11, 2008


I don't think spiderwire is saying this post is Pepsi Blue, rather making a joke that the "news" is vulnerable to malicious pollution as well, with a 'bad' in the midst of the Pepsi.

My Coca-Cola options have made me rich; thanks MetaFilter!
posted by spiderwire at 11:34 AM on September 12, 2008


« Older DC Comics' All-Star Batman and Robin series (from ...  |  Harper has called a federal el... Newer »


This thread has been archived and is closed to new comments