This has been around forever. I wrote a research paper on using microblogging services to predict political events about a half decade ago.
Three years ago my friend wrote his senior thesis on using it to predict stock market pricing. Brilliant guy, Microsoft Research now, and the conclusion was that once you factor in bad entity resolution (AAPL vs "this apple is delicious") and account for transactions fees your performance more or less matches the market.
He was WAY too early in 2012 to achieve any sort of statistical significance in terms of number of traders talking online consistently. I have a client running 75 different strategies just on our data alone and there are quite possibly hundreds more, not all of which involve transactions. The ability to predict volatility is one that comes to mind.
2 comments
[ 3.5 ms ] story [ 17.8 ms ] threadThree years ago my friend wrote his senior thesis on using it to predict stock market pricing. Brilliant guy, Microsoft Research now, and the conclusion was that once you factor in bad entity resolution (AAPL vs "this apple is delicious") and account for transactions fees your performance more or less matches the market.