AnalyticBridge
A Data Science Central Community
Your Tweets Could be Worth Millions, as Data Mining Algorithms Analyze them to Predict the Stock Market
London-based hedge fund Derwent Capital Markets said it had successfully marketed a new venture to a series of high-net worth clients that makes investment choices using information gathered from over 100 million daily tweets.
Simply put: the fund mines the Twitter-verse to gauge market sentiment, and that information-which the firm futuristically brands as "The 4th Dimension" is used to drive the portfolio's holdings.
The 'Twitter Fund', officially marketed to clients as the Derwent Absolute Return Fund, has already attracted at least £25 million in investments, according to the fund's manager Paul Hawtin-who is also the firm's founder. And it is currently in discussions to hire John Bollen, an Indiana University professor who has championed academic theories linking market performance to Twitter moods.
In all, it's a bold strategy, with even bolder profit expectations.
Hawtin is confident he can achieve annual returns of 15-20 percent for the fund. But he hopes that's just the beginning. "I believe with the Twitter indicator (once fully optimized) we can achieve even better annual returns," he said in an emailed statement.
But even if the fund doesn't return a dime on its money, it may have already made history. Hawtin believes the Derwent Absolute Return Fund is the first ever "to use real time mood analysis as a major part of the investment decision process."
An informal search of publicly available hedge fund strategies seems to confirm that.
Asking investors to park millions in an investment vehicle that relies on the random musings of 190-plus million Twitter accounts was not without its challenges.
Hawtin says that investors were initially worried that the fund was simply going long or short based on its Twitter analysis. However, he says it's a more nuanced process than that. "Once they understand it's a far more sophisticated system they then realize the potential," he said.
Whether or not the fund can return on its lofty profit promises remains to be seen, but whatever the fund's future may be, it is sure to open the door for a new round of quantitative experimentation-as the investment world looks for new ways to profit from the endlessly vast amount of information streaming to blogs, Facebook, Twitter, and other social networking sites every second of every day.
Welcome to the 4th dimension.
Source: http://finance.yahoo.com/news/Your-Tweets-Could-be-Worth-cnbc-45801...
Comment
-
Comment by Patrick Stewart on January 10, 2011 at 11:01pm
-
"Twitter-verse". Nice.
HP has actually been mining the social media world for a few years to predict the opening weekend sales of movies. One recent title was predicted to be a blockbuster by media analysts. HP said it was only going to make 28.8 million. The result? 28.5 million. Far fewer people will be tweeting about hedge funds than movies, but it is a proven concept.
-
Comment by Vincent Granville on January 8, 2011 at 8:18pm
-
Tom -- There are many people who are terrible at predicting the stock market, who have no investment knowledge, and are consistently wrong (say 70% of the time) in their predictions. As long as they are consistently wrong, using their forecasts (changing their 'buy' signals to 'sell') is better than using forecasts from great sophisticated investors who are correct only 60% of the time.
The stock market typically works in a way that is counter-intuitive, and behavioral trading is what kills the novice. Exploiting the novice's ignorance (via his/her tweets) could indeed work well.
Similarly, exploiting pump-and-dump price alerts scams (email spam that you get in your mailbox or by fax) can work very well, if you change the recommended action from "buy" to "sell", assuming you wait long enough before receiving the spam and short-selling the recommended stock (indeed, you can do statistical analysis to determine the optimal time lag between receiving the spam and proceeding to a short sale).
-
-
It depends how the data is used. You can cluster Twitter users in 3 categories:
- those who usually make correct predictions
- those who usually make wrong predictions (sell when they say buy)
- those who make both wrong and right predictions (you can't use their tweets as predictors)
You can also be more granular, e.g. identify users who usually make correct predictions but only for some categories of stocks. Then your leverage their tweets, but only for these particular stocks. You can also gather info about the twitters (e.g. education level if available in profile) and create a mathematical model that tells you how to leverage tweets based on education level / location / gender etc. And you should also compute the optimum time lag between when a tweet is made, and when it can be used to buy / sell stocks.
-
Comment by Ralph Winters on January 7, 2011 at 1:28pm
-
This seems like a pretty irresponsible thing for a hedge fund to do, but I'm not surprised. Do we really need more junk science? This also runs contrary to tradition wall street sentiment indicators which say the best time to be in the market is when you are ultra-pessimistic.
I've been trying to track down the original "research study" with no luck. So without that, I will judge it to have about the same amount of predictive power as handwriting analysis.
-Ralph Winters
© 2020 AnalyticBridge.com is a subsidiary and dedicated channel of Data Science Central LLC
Powered by
Badges | Report an Issue | Privacy Policy | Terms of Service
You need to be a member of AnalyticBridge to add comments!
Join AnalyticBridge