Computers Get Sentimental for Traders
By TOM GROENFELDT
Traders get plenty of pricing information; Exegy, which runs a ticker
plant in a data center outside New York City reported 6.65 million
messages per second during an October peak. Several vendors offer
high-speed quantitative tools to analyze such data. Now, in addition to
prices, traders can see where it might lead with Recorded Future and get
a feel for the markets through a Thomson Reuters sentiment feed.
Recorded Future can predict at least some aspects of the future by
monitoring the equivalent of 300,000 pages per hour from 40,000 to
50,000 Internet sources and it can figure out what the text means.
Steffan Truve, co-founder and chief scientist at the company suggested
asking Google what is happening in Stockholm next week if you want to
see how Recorded Future differs from a search engine.
The two-year old Massachusetts-based firm is partly funded by Google
and partly by the US. government’s In-Q-Tel, which makes investments to
benefit the United States intelligence community.
Recorded Future has already helped traders to choose which stocks to buy
and sell. By collecting information and sentiment, it can pick the 50
most popular and the 50 least popular, said Christopher Ahlberg, the
company’s co-founder and CEO. By going long on the former and short on
the latter, traders can make money.
“Organizing the news and companies by decile (dividing them into
groups of ten), you can prove the top decile significantly outperforms
the bottom. This can turn out to be a very nice trading strategy, either
as a stand-alone factor or incorporating the data in an overlay to an
existing model.” From May to August the Recorded Future strategy had a
10 percent gain against a 10 percent loss for the S&P 500.
Recorded Future works with structure such as people, places,
products and companies; events such as meetings, travels, acquisitions,
earning calls and natural disasters; and ontologies or hierarchies that
explain groupings such as world leaders, corporations or technology
areas. And the company’s analysis treats time in several ways—when an
event was reported and when something occurred or is expected to occur.
It also measures momentum or media buzz.
It’s worth taking 10 or 15 minutes buzzing around the company’s Web
site to see some examples and to play with the views using sliders on
the Web pages.
Steve Holden, Recorded Future’s community manager, said that in an
areas like competitive analysis, a user could learn about Zynga’s
business relationships over the last two years in a few minutes.
Recorded Future isn’t the only company trying to get a feel for the
market that goes beyond pricing data. Thomson Reuters has developed a
data feed which reads the news, measures sentiment with sophisticated
linguistic analytics and distributes the results to both algorithmic
trading engines and real live human traders.
“We filter out the noise on the Internet and concentrate on the most
relevant, impactful and market-moving news,” explained Richard Brown,
head of quantitative and event-driven trading solutions at Thomson
Reuters. “That allows you to make sense of petabytes (1,024 terabytes)
... and get to what you care about the most. Our analysis techniques,
which look at what company, what context and sentiment, all help you to
be able to take sips from the firehose.”
High frequency trading (HFT) engines can take sentiment data feeds
based on automated analysis of news and electronic media along with
market data and news from Thomson Reuters. Behind the sentiment is a lot
of powerful processing and recently developed skills in reading news by
computer and understanding it. The feed is even able to understand
sarcasm.
“We have no shortage of interest, and we are seeing significant
take-up in the number of clients who are using it,” added Brown. “Most
of them are using it for quantitative trading strategies, but the
intelligence added to the text enables a number of ways for humans to
use it more intelligently to filter out the noise and get a broader view
of the company, sector, or market.”
Published: February 12, 2012
Issue: February 2012 Issue
Comments
applications for traders?
Very interesting and well presented. Does this technology have application to trading, e.g. binary derivates?
Marvin K., Chicago, Mar-02-2012