London - Amid all the hype surrounding news and social media sentiment indicators, one question comes up frequently: Is anyone making money by mining these information sources?

For one group of investors at a Battle of the Quants conference in London, the answer was a resounding "Yes".

"Sentiment is, perhaps, the most important thing in markets," said Richard Edwards, managing director at HED Capital Management.

"Definitely it works, there's no doubt about it," said Raul Andreas Glavan, the winner of last year's IronTrader competition in Germany.

Otto Van Hemert, a quantitative portfolio manager at IMC Asset Management, said his fund in managed futures is already up 9 percent for the year using sentiment analysis techniques.

One after another, the panel members lined up to talk about how they had found success in this area. One who captured the audience's attention was Glavan. who went so far as to describe in detail one of his strategies.

Glavan said he felt free to reveal a technique he had used with Twitter because it was not scalable. His model identifies tweeters who have developed followings and who, history had shown, created enough market buzz to move shares higher.

By filtering the Twitter sphere correctly, Glavan spoke of how he was able to get in trades based on their tweets and the responses to them. He would also ride the market's eventual turn lower, he said, by shorting the stocks. This, he said, worked 70 percent of the time.

"There's no sense for me to trade the news because there are a lot of people who have their computers inside the stock exchange colocated. I cannot beat them, they have more speed, they have more research," Glavan said. But, if he had tradable information before the rest of the market -- such as the insights gleaned from his Twitter tracking -- "then it's easy to trade".

Glavan said these techniques had value now because of their rarity. In 10 years' time a lot of that value may have evaporated, but for now he believes social data mining has a lot of mileage.

"At the moment, it's really unexplored," Glavan said. "Social media, it's new and you can have an edge at the moment."

Edwards of HED Capital said sentiment trading worked because of feedback loops.
"You see this in markets in particular, where positive feedback leads to a self-sustained trend," he said.

Rising prices attract more buyers; then, instead of leading to price equilibrium because of additional supply, the opposite occurred as sellers held back in the expectation of still higher prices. Of course, there was the exact same effect on the way down.

A Range of Sources

Van Hemert said IMC uses a range of information sources, everything from standard market data to Google trends to news measures. Randy Saaf, CEO of Alpha Genius, also saw value in social data.

"When we're thinking about sentiment and news, news tends to be big events. You know, people don't write about pedestrian things. They write about big, interesting scandals. Whereas Twitter and blogs, they can be pretty pedestrian - and there's value there," he said.

Edwards spoke of news and other kinds of information as part of one continuum.
"We think that news and opinion are just data points in the landscape of the evolving view of the market. And we think that sentiment is just that, it's the mass mood of the market at any one time, so that's why at tops you have such a bullish crescendo and at bottoms you have such a torrent of bearishness."

Edwards added: "So we think measuring, using the news, is a very good way of identifying data points in which you can define the story as it unfolds. And social media is just a faster and newer and arguably better way of doing that."

Saaf of Alpha Genius said the nature of sentiment trading differed from the "hardware battle" that could be seen in the high-frequency trading arena.

"This is about getting the signal from the noise," Saaf said. "The successful one .. is going to be the one who continues to turn the screw on pulling that signal from the news and the noise, as more and more people start to trade sentiment."

In such a trading strategy, latency was less an issue. "This is not a speed question. It's an analysing-big-data question," he said.

Sven Steude, a professor at the University of Zurich who also took part in the panel, noted that there were still vast terrains of information that could be tapped. He highlighted all the streaming video content online waiting to be analysed.