Wednesday, May 18, 2011

Twitter Hedge Fund

The World's First Twitter-Based Hedge Fund:

This week, Derwent Capital Markets, a London investment firm, launched a $40 million hedge fund that will use Twitter to guide its investments. The world's first social media-based hedge fund will monitor a selection of tweets in real-time to feel out market sentiment before placing its bets...
The professors harvested tweets for key words and plugged them into an algorithm to determine the mood of the broader market. Using this mood index, the professors predicted the Dow's daily fluctuations in 2008 with an astounding 87 percent accuracy.

I remember smart people looking at keyword algorithms to predict asset fluctuations in the 1990's, and these projects invariably foundered. Thus, it's not a new idea, just a new medium, in this case Twitter. I'm skeptical.

4 comments:

Anonymous #5 said...

Yeah, if you read the paper this is supposedly based on, they predicted the direction of the stock market in 13 out of 15 days in December 2008. Of course a simple "reversion" method would have gotten 11 of those 15 days right.

It's been a while since I read up on this, but I'm pretty sure if you added another variable to their preferred model, they would have only gotten less than 50% of the days right, which doesn't make their work seem very robust.

Mercury said...

Yeah, and how did it do in 2009 and 2010 at predicting Dow fluctuations?

I suppose Eric that it could be argued that Twitter data is more of a fine-grained, primary source than whatever was being used in the 90's....but the premise still has to be that retail or even human sentiment (as opposed to The Fed, HFT algos or whatever)is the determining factor driving the Dow these days and I'm not sure that's the case right now.

Anonymous said...

Over-fitting too: word recognition is basically highly multivariate time series (with second moment effects to boot!). That gives too much leverage to whomever makes the analysis to cast whatever results in a good light. I'm sure if they were to explicitly spell out the matrix of the 100 best source*expressions you would find some combinations that would be very rarely visited going forward...

Anonymous said...

I had to laugh when I read that people are actually giving money to those illusionists.

Some will never learn.