Seeing into the future with Twitter: a brief history, and a new approach

paper just released by the University of California adds to the noise around Twitter as a stock-picking tool. It focuses on analysing the relationship between stock price movements and Twitter metrics on related tweets such as number of hash-tags, number of re-tweets, number of followers for the user that posted the tweet, and so on.

Twitter and other microblogging services provide an efficient way for people to publish their opinions informally in real time. And people tend to tweet ideas very soon after having them. So it’s not surprising that analysts seized on Twitter early on as a possible leading indicator of market sentiment.

Here’s a quick history of this field:

  • October 2010: A team at Cornell University performs a pioneering analysis of the correlation between tweet sentiment to see if it can predict the closing price of the Dow, and finds that it can — with a claimed 87.6% accuracy in predicting whether it will move up or down. They use a clever self-organizing fuzzy neural network to “learn” the correlations between tweets and outcomes.
  • November 2010: The Technical University of Munich (TUM) follows this up by analysing 250,000 Twitter messages and finds a correlation between Twitter sentiment and abnormal stock returns across a range of stocks. They claim that an investor acting on this information would have achieved an average return of “up to 15%” over the six month period that they analysed. Their approach uses sophisticated computational linguistics to perform semantic analysis of the posts. Crucially, they find that users providing above average investment advice are retweeted more than others, providing a handy ranking algorithm.
  • April 2011: Economist Tim Sprenger of TUM sets up, a website to aggregate stock-related social media content in real time.
  • May 2011: Derwent Capital Markets launches what it claims is Europe’s first hedge fund to utilise sentiment drawn from Twitter analysis. (Previous Caplin blog post on Derwent Capital here)
  • August 2011: Modulus Informatics launches, free site that lets users pick stocks to analyse before its system pulls in Twitter data and comes up with buy or sell signals.
  • January 2012: Microblogging search specialist Topsy Labs announces that is developing a Twitter sentiment analysis tool for stock market investors, to be launched this year, after claiming market-beating predictions of movements in Netflix and Apple stock.

The team at Bourns College of Engineering at the University of California Riverside has now performed what may be the most rigorous analysis to date. They examined the entire Twitter dataset over the first six months of last year, searched for mentions of a sample list of 150 S&P500 companies, and came up with over 26 million tweets. They then looked at the correlation between stock price movements and trading volume over the days before and after the tweets with a range of different metrics.

Their conclusions? The best metric is “connectedness” (read the paper if you really want to know exactly what the means), and there is a correlation between tweet volumes and trade volumes (not terrifically surprising) but only a weak one with price movements. Yawn.

Can Twitter sentiment make you rich? One thing’s for certain: if Twitter analysis really can be used to predict the market, hedge funds and investors will quickly pile in and arbitrage it until it doesn’t work any more.

So if you think you know how to do it, hurry up.

One Response

  1. […] Seeing into the future with Twitter – a brief history, and a new approach: This post stands out because the author has obviously covered the very recent issue of trading based on Twitter sentiment right from the beginning – his historical outline of the key events is testament to that. He also points out that the latest developments cover not sentiment-based trading, but the correlation between tweet volumes and trade volumes. His take on this space is negative but this is justified as he’s done the work to come to this justified conclusion! Share this:TwitterFacebookPinterestDiggStumbleUponEmailTumblrLinkedInRedditPrintLike this:LikeBe the first to like this post. […]

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

%d bloggers like this: