Social media may be here to stay, but the chatter surrounding the closure of the Twitter-based hedge fund operated by Derwent Capital Markets shows the platform is in its infancy as a buy-side trading tool.
Under normal circumstances, a hedge fund with a paltry $40 million in assets under management wouldn't get a second look from the media or the investment community. But the world's eyes were fixed on Derwent Capital's Absolute Return fund from Day One, since the hedge fund's strategy was built solely on an algorithm that used Twitter to predict the direction of the stock market.
The algorithm, crafted by computer scientists at Indiana University and the University of Manchester, was designed to scan and analyze large Twitter feeds on a daily basis in order to capture the pulse of the public's mood. It would then mine that data to detect where the stock market was headed three or four days in advance -- to the tune of an 88 percent accuracy rate, according to Indiana University professor Johan Bollen, who helped build the platform.
[How Useful is Social Media-Based Sentiment Analysis to the Buy Side?.]
During its lone month of trading in July 2011, the algorithm proved to be a success, earning a sterling return of 1.86 percent, besting the average hedge fund and the broader market for that period. Nevertheless, Derwent chose to shutter the hedge fund and reportedly decided -- at the urging of one of its largest backers -- to sell the platform as a tool for private investors.
Not Ready for Prime Time
Yet whenever Derwent's Twitter-based algorithm hits the marketplace, it's unlikely that buy-side trading firms will be lining up to add it to their arsenals. Experts say that while such platforms hold considerable promise as trading tools for hedge funds and traditional asset managers, it'll be years before they catch on with the typical buy-side trading desk.
"Most traders would scorn this," says John Bates, the founder and chief technology officer of Progress Software. "I remember when this came out -- I got an email from one of our customers saying he'd rather put his private parts in a guillotine than trade on Twitter."
According to Bates, the addition of tools like Derwent's to trading desks will likely follow the same, slow path that's currently being tread by news-based analytics systems such as those provided by Dow Jones and Thomson Reuters. And even those systems aren't all that widespread on the buy side, he notes.
When news analytics first appeared, Bates says, there were traders -- some of whom were his firm's clients -- who traded on news with a sort of fuzzy logic approach. But once firms like Dow Jones and Thomson Reuters began putting tags in news, it evolved into a type of structured market data product.
"There are a few firms that use that in their strategies -- I don't think it's that widespread, and there are even fewer people using Twitter as a trading signal," Bates explains. "We've just got to find the killer application that [social media] is useful for. Right now I think it's just one indicator that might be useful for things like finding the bottom of a market or determining whether people feel good or bad about the economy."
Derwent, meanwhile, isn't the only firm to seek a profit by trading on, or selling, a social media sentiment analysis system. Santa Monica, Calif.-based MarketPsy Capital also launched, and later closed, a hedge fund that used social media cues to formulate trades. And like Derwent, MarketPsy is now selling its sentiment analysis tool, though the Financial Times reports that the firm also is gearing up to launch a new private fund this year after making some adjustments to its algorithm.
Even if traders were ready for tools like these, it will take years for buy-side side firms to implement them properly. David Polen, Fidessa's head of business development, says before a firm could adopt this type of technology, it would first need to be equipped with a research engine that could help determine how valuable the information actually is, since much of it is merely noise.
"Whether you're talking about social media or news releases, the missing component of the puzzle is actually tying this into your research engine," Polen explains. "Until the research engines contain correlated information between quotes and trades, sentiment-based and low-latency news, you can't answer that question. If it turns out there's a really bad noise-to-signal ratio with this, these tools won't be indispensable."