Tracking tweets proves profitable for Derwent Capital
A hedge fund that uses Twitter data to drive its trading strategy returned 1.85% in its first month of trading, according to an investor in the fund, in the first sign that social media data can be used successfully to enhance electronic trading techniques.
Derwent Capital, which finished its first month of trading at the end of July, beat the S&P 500 which fell 2.2% in July, while the average hedge fund made 0.76%, according to Hedge Fund Research.
The fund, which declined to comment, uses sentiment data mined from millions of Twitter messages, or ‘tweets’, to predict market movements. The strategy is based on research published by the University of Manchester and Indiana University in October which demonstrated that the number of emotional words on Twitter could be used to predict daily moves in the Dow Jones Industrial Average.
Derwent Capital scans a selected 10% of available tweets at random and will then categorise these messages into one of a range of mood states, which could include 'alert', 'vital' or 'happy' from which the firm's technology will make predictions about potential stock movements. The initial research showed that the algorithm predicted movements in liquid stocks with 88% accuracy.
Although it is early days for the fund, July's performance provides the first indication that the unstructured data provided by Twitter can be successfully used to feed trading algorithms. The results may go some way to convincing sceptics who have argued that the five-year old Twitter's relative youth and its unstructured, unedited nature, makes its data unreliable as a sentiment tool.
Speaking to Financial News in early March, however, Paul Hawtin, founder of Derwent Capital, said he was “very confident” about the strategy. He said: “We watch to see how each of the different mood states changes in real time. The biggest sentiment change is reflected in the market around two to four days later.” He added that it is the unprecedented global nature of Twitter that defines its use as a sentiment tool.
The Twitter service allows users to push out messages or 'tweets' of 140 characters or fewer to a base of subscribers -- known as 'followers' -- who are then able to respond through their own Twitter account. The messages appear as a real-time list of rolling messages on the user's Twitter profile page. To date, the service has more than 200 million users.