Last Thursday, we hosted a webinar – “Trading on Macroeconomic News” with Need to Know News (NTKN), a subsidiary of Deutsche Bourse Group Company. With pre-built connectivity to NTKN’s AlphaFlash, we demonstrated how a trading firm could easily incorporate low-latency news algo data feeds into their trading systems. In the 50-minute webinar, we built a basic algo that will trade EUR/USD based on the data of non-farm payroll employment report.
Trading on unstructured or non-traditional data is a topic that is gaining more and more attention. Adam Honoré (@adhonore) at Aite Group reported that 35% of quant firms are now exploring machine-readable news feeds in some way, compared to only 2% in 2008. The development of electronic trading and explosion in data volumes over the past decade play key roles in this phenomenon. Yet, most market participants are still trying to figure out effective and efficient ways of transforming unstructured data into increased profits or prevented losses.
One fund, Derwent Capital has reported success already, announcing a 2% profit from its $40 million dollar fund, with investment strategies based on Twitter mood - derived from a paper published by Johan Bollen and Huina Mao at Indiana University, Bloomington, and Xiao-Jun Zeng at the University of Manchester. In 2009, Zhi Da and Pengjie Gao at Notre Dame and Joseph Engelberg of UNC also proposed correlation between Google Search Volume Index (SIV) and stock prices.
Here at StreamBase, we were one of the first financial technology firms to offer clients a way to connect to unstructured social media content, such as Twitter. Media interest in this area has been high, with coverage from CNBC, the FT, the WSJ, and USA Today; though enabling customers to trading on social media, is really only a small percentage of what we do. The scalability and flexibility of StreamBase CEP makes it a great platform to handle unstructured data in real-time.
Leading economic indicators, such as GDP, consumer confidence index, interest rates and unemployment reports are commonly used for signal generation but a firm needs to make extra efforts to really gain a competitive edge with unstructured data. Simply categorizing news into positive, neutral or negative might not be sufficient enough for effective algorithmic trading. In 2010 we demonstrated how StreamBase can be used with a LingPipe, a natural language processing tool kit that can understand the implications of a news story vs. market expectation.
Although it is rare to see signal generation or alpha seeking strategies solely based on news events, some sophisticated firms incorporate unstructured data and sentiment analysis to develop algos that react to market volatility after a particular news item breaks.
Processing of unstructured data has relevance not only for alpha-generating strategies but also for risk managers and regulators who could use machine-readable news and sentiment analysis to simulate a firms’ exposure or market fairness under extreme events, for example, the recent downgrade of government credit ratings.
Trading on unstructured data will continue to be a popular topic. If you have any new ideas for using other non-traditional data for trading, please let us know. Below are some additional links which might prove interesting.
Further reading:
Using Twitter to Predict Stock Market Swings, New York Times
Twitter Mood Predicts The Stock Market, Technology Review
Hedge Fund Will Track Twitter to Predict Stock Moves, Bloomberg
Which Stocks Will Rise? Ask Google, SmartMoney
In Search of Attention, Zhi Da, Joseph Engelberg and Pengjie Gao
Computers That Trade on the News, The New York Times
On IBM, Unstructured Data, and CEP, StreamBase Event Processing Blog
Twitter for those dirty, alpha-seeking strategies, FT Alphaville
