You’ve probably seen this before: A company announces some kind of bad news and the stock price tanks. This happened to CVS few days ago when Walgreens announced they were ending a partnership with CVS. The price of CVS stock dropped something like 10%. That was big news and it triggered a number of people to sell their ownership of CVS, making the price go down.  Of course their relationship was repaired and the stock price bounced back.

But sometimes the news is less significant. It might just be something small like a new president of a division, or the launch of a new product. These announcements happen all the time, and don’t seem to have an observable impact on the stock price in the short run, at least the same way something big does like CVS. But these days on Wall Street, it’s not about making 10% in a day off some significant news, it’s about making fractions of a percent dozens of times a day. The buying and selling happens so fast, that a human could never make the moves fast enough. But, where there’s a will there’s a way nerd with some programming skills that designs a system to capitalize on these fluctuations.  And they have shown they can make good money by doing it.

And those nerds (called ‘Quants’ on Wall Street) are Robert Shumaker and Hsinchun Chen, and they think they’ve figure out a way to make money on even the slightest news item. Through a process that is probably best described in a Professor Frink voice and makes Einstein roll in his grave for wasted brainpower, they’ve identified key words in news stories that allow them to predict temporary price moves in a stock. They have to act fast, within about 20 minutes, and the moves are only a few pennies. But in typical professor fashion, they tested it against some of the biggest quant funds on Wall Street and it works. The way it works is their program scans news stories and looks for key words. Certain words have a bigger impact on price than others. Depending on the words coming in to the algorithm, the program will either short a stock or buy it.

So the question then is, what words make the algorithm short a stock and what words make it buy? You couldn’t guess them even with 1000 tries.

Buy: planted, announcing, front, smaller, crude
Short: hereto, comparable, charge, summit, green

Why these words? I can explain some. ‘Announcing’ makes sense, companies make announcements all the time; a good announcement could mean an increase in buying. ‘Charge’ also makes sense, as it could mean a criminal investigation, or an additional expense that eats into profits. And I guess investors don’t like it when companies go ‘green’. But I can’t really explain any of the other ones.

But like any computer system, it potentially has flaws. Flaws that caused the flash crash a few weeks ago. Computers can’t reason, so they can’t tell if news is legitimate or not. In 2008 a newspaper accidentally ran a story from 2002 about a United Airlines bankruptcy. It sparked a huge selloff and investors lost a ton of money. An enterprising hacker could launch a PR release campaign that causes this algorithm to trigger buys or sells. I’m not saying this system isn’t sophisticated, but there isn’t a system on the planet that can’t be tricked or hacked.

This kind of develop sure is interesting, and I think there’s a place for it in the future. Maybe if I have $1 million to invest I’d put $50k in a fund that does this. But I’d never trust all my money to a computer; not in a world with flash crashes, hackers, and humans writing code.

Via The Reformed Broker

By The Weakonomist

The Weakonomist is the founder of Weakonomics. Clicking the picture of the puppy will tell you more about him. If you like what you're reading consider subscribing to the blog, following him on Twitter, and checking out the related posts below. Please consider a donation if you really enjoy Weakonomics. Also keep in mind that nothing I say can be construed as financial advice.

categories: investing