Welcome to the Future Market
The year is 2011, and sophisticated AIs carefully watch trends and track/correlate/analyze vast amounts of data to inform sophisticated maneuvers in the global markets.
Or, well, maybe a bunch of dodgy scripts are doing clumsy sentiment analysis on various volatile and imprecise social networks and rumormongering news services, and they can’t tell the difference between Anne Hathaway and Berkshire Hathaway.
23 Mar 2011 at 00:44
Of course, if enough hedge funds are doing this, then there may be an exploitable correlation between Anne Hathaway and Bershire stock prices.
The interesting thing here is what r values aren’t profitable. Is it just -.1<r<.1? Or could it be narrower?
I like how the article doesn't answer the "why" question by positing cause. It's all correlation. The traditional way we'd use "why" no longer matters when you have terabytes of data to analyze.
So maybe r=.03 can be profitable.
23 Mar 2011 at 13:21
This article explores the theoretical possibility of trading algorithms that try to game weak correlations between ostensibly unrelated events/things, but it doesn’t prove that such a trading bot exists, much less account for the seeming correlation between the Sage of Omaha’s company and a certain toothy actress.
Indeed, how do we know that it’s really the actress that is Buffetted by the fickle winds of the Internet? What if this really about William Shakespeare’s wife?
Regardless, I’d like to see some statistical analysis of the significance of this correlation. Some p-values would be nice, or failing that, some null-hypothesis testing involving Jennifer Aniston.