Can Twitter Tell You When to Buy and When to Sell?

2 7 Loading

There are two central drivers of stock price demand—fundamentals (sales, revenues, profits, etc.) and how investors feel about fundamentals (sentiment). Sentiment tends to erratically drive short-term pricing, while the longer cycles move on fundamentals. If you talk to a buy-and-hold investor, like Warren Buffett, he will tell you short-term investing (day trading, for example) is a fool’s game—there is no predicting sentiment.

But Derwent Capital Management (DCM) thinks that may have been true, once, in ancient times before information technology enabled social networks. But now there is a wealth of hard data on real time sentiment. All one must do is set up an algorithm to mine it, process it, put it on a scale—in this case from 0 to 100—and sell it to retail investors.

And that’s exactly what they’ve done.

Wondering what your favorite stock (or currency pair or commodity) is about to do? You need merely check the the DCM trading platform’s Twitter indicator. The burning question is, does it work?

DCM thinks Twitter can tell us how investors feel about stocks and help predict prices.

Maybe. Sometimes. Probably better for the broad market than individual stocks. But first and foremost—with all things financial, do your homework. In this case, despite whatever studies say it works (and there have been some) Twitter sentiment analysis has a short track record, and correlation does not equal causation. Beware.

Second, for measurements of Twitter sentiment to be useful or accurate for specific stocks, you need the right people Tweeting enough at the right time. And third, the stock market is reflexive—a tool only works so long as you’re one of the few people utilizing it.

Twitter has 500 million users, give or take. Of those 500 million, far fewer are active members (log in periodically)—and fewer still regularly send Tweets. For Twitter sentiment to be a useful barometer, you needn’t require Tweeters be professional investors. But you do need them to actually care enough about stocks, commodities, or currency trading to Tweet key words about them. How many actually do that isn’t clear. The number is far from zero—but is it enough to be meaningful?

The accuracy of Twitter sentiment will depend on how many Tweets a given stock, commodity, or currency pair receives when the investor wants to test sentiment. One can imagine a scenario when a huge stock is in the news and gets sufficient Tweets to form an interesting sentiment picture. But will the same be true of the other 250 UK stocks currently covered? All the time? And what about the 8,000 additional global equities they want to cover in the future?

Ultimately, from moment to moment and investment to investment, the tool’s accuracy is likely to vary. Maybe a lot. To solve this problem, the firm could add numerical strength representations of each rating (how many Tweets back it up, the relevancy of those Tweets, etc.). This would help investors choose when and when not to employ the tool.

But further complicating matters is the direction of sentiment in general—does it look forward or backward? One oft quoted guage of sentiment is consumer confidence. The problem with consumer confidence is it usually reflects how folks are feeling about what’s happened in the recent past. Yet it is largely uncorrelated to how they’ll act in the future. US consumer confidence hit all-time lows just before the stock market bottomed in early March 2009—six months on, US stocks had posted 50% gains.

Finally, Twitter sentiment’s value as an investment tool will vary indirectly with its popularity. The more predictive Twitter sentiment proves to be, the more investors will use it. And the more people act on Twitter sentiment algorithms (plural, there are already competitors and if it appears to help some—more will jump in), the faster the market will discount them. Whatever advantage the tool initially provided will disappear.

Ultimately, gauging sentiment is an incredibly difficult nut to crack. Aggregate social systems like the stock market or economy are among the most complex systems in existence. If we can’t forecast the weather more than ten days out—how can we expect to predict millions of irrational brains playing off one another?

Maybe future AIs and supercomputers will accomplish more than is imaginable today. Until then—at most, Twitter and other social network data should be combined with other rigorous analysis before pulling the trigger on a trade.

Image Credit (top to bottom): Andreas Eldh, Manuel Bahamondez, David Paul Ohmer

 

Discussion — 7 Responses

  • dobermanmacleod January 29, 2013 on 1:25 pm

    There is a certain irony that an article about Twitter giving stock tips would give you a stock tip of the century. There is a mega-trend now occurring that you ought to be aware of, which could make you rich (I have no vested interest financially, but the urge to inform you):

    http://coldfusionnow.org/is-it-finally-happening-a-big-market-shift-in-the-energy-sector-is-claimed/

    http://coldfusionnow.org/is-it-finally-happening-supporting-evidence-for-a-big-shift-in-the-energy-market-caused-by-lenr/

    LENR is going to emerge onto the market this year. Leonardo, Defkalion, or Brillouin seem like the best candidates.

  • jauker January 29, 2013 on 3:43 pm

    The correlation of sentiment analysis of twitter predicting stock market moves, also has been shown to exist with the blogger platform live journal (http://comp.social.gatech.edu/papers/icwsm10.worry.gilbert.pdf).

    It has been suggested by some that this is because social mood is reflected in both tweets and stock indexes. Several books have been written on this topic and this hypothesis has been around since at least the late seventies. This hypothesis is referred to as “Socionomics” and there is a lot of interesting implications and exciting research in this area. According to the Socionomic hypothesis social mood not only influences social media and equity prices, but also other social events ranging from wars, to epidemics, to elections, to fashion just to name a few.

    This hypothesis is still way outside of mainstream finance, but the hypothesis is growing in popularity. Some of the assertions put out by the Socionomic Institute seem outlandishly speculative, but recently there have been some compelling developments. If you have any questions feel free to ask me because I’ve spent an unhealthy amount of time reading about this topic.

  • senti-men-t February 3, 2013 on 9:51 am

    Removed from whether it works or not; it is very much in the interest of any trading provider for the sentiment to be inaccurate… to employ sentiment in the context of a trading platform means either one of a few things – that sentiment is occasionally correct, purely by chance – or that sentiment is indeed accurately able to predict market movement, ultimately resulting in the provider losing more money than they make – or that the sentiment is falsified in an attempt to persuade users to bet in the wrong direction (though this would mean that somehow market movement could be predicted, which is cannot, and is the very reason why sentiment is carries absolutely no substance).

    • dobermanmacleod senti-men-t February 3, 2013 on 9:32 pm

      I disagree. “Sentiment” is actually the main method of valuation. Winning or losing on the market is driven more by fluctuations in value based upon mod sentiment than by solid evaluations like dividends or net worth of the company. If you can read “sentiment,” then you can judge trends in valuation, thus being able to make money on the volatility.

      • senti-men-t dobermanmacleod February 5, 2013 on 10:28 am

        utter bollocks

      • senti-men-t dobermanmacleod February 5, 2013 on 10:29 am

        people who say ‘I disagree’ a typically idiotic fools who think they’re clever…

        • dobermanmacleod senti-men-t February 6, 2013 on 12:48 am

          Ironic – I think that people that say “typically idiotic fools” are typically projecting.