October 9, 2018
News-driven alternative trading strategies have been slower to take off in Asia than in the West, but that’s changing. A survey of attendees across Bloomberg’s Machine Learning Decoded events in Sydney, Singapore, Tokyo and Mumbai, poll respondents were almost evenly split between machine learning users and non-users.
Of all respondents, 42% are considering how to incorporate tools into their workflow and 31% are in the process of applying machine learning to their operations.
There are now approximately 2 million headlines a day that pass across the Bloomberg Terminal, which can be overwhelming. To take advantage of the information distributed in this sea of unstructured data, firms are looking to consume the text of news in machine readable form and enable their trading strategies to take advantage of the signals hidden therein.
In the Bloomberg survey, the majority of participants are using real-time data (45% in Sydney, 43% in Japan, 35% in Singapore), followed by fundamentals (financial results, securities prices), news and social media sentiment.
Systematic strategies must react to huge volumes of market-moving news and data in order to generate alpha. Bloomberg’s Event-Driven Feeds solution provides real-time, highly structured machine-readable news, coupled with news analytics to allow the user to detect spikes in readership, publication flow and sentiment. It comes with accurate, granular metadata. News sentiment is scored at both individual story and corporate level using a proprietary Bloomberg algorithm.
The challenge is finding correlations between what’s being said and what the market does. An example may be oil traders looking for spikes in news reports mentioning the term “bear market”. Spotting a surge early can be a signal to sell. When such conversation nears a peak, it can be an indicator to buy crude to get ahead of the curve.
To be successful, data quality is critical. For prop desks, market makers, quants or high frequency event-driven traders who rely on non-display trading applications (black or grey boxes), every news story represents a critical opportunity that must be acted on quickly and accurately. A deep news data archive is essential to enable meaningful backtesting of new strategies.
To date, firms with news-driven strategies have mostly used analytics to evaluate stocks. This is expanding into commodities and foreign exchange. Analyzing equities is more straightforward given the more structured, repetitive nature to a lot corporate news, like earnings. Currencies have far more levers, including politics, monetary policy, economic data, money flows and disasters.
For busy fund managers tracking large stock portfolios, the Bloomberg Terminal has tools for identifying companies and topics that are attracting heightened interest from the financial community. TREN <GO> lets you see the topics and companies that Bloomberg users are increasingly interested in reading about, surges in news coverage, and those with the most positive and negative news sentiment. It also identifies surges in social media interest.
For more news insights, users can access GN <GO> for news publication heat and sentiments against a share price. GT <GO> on the other hand, displays social media heat and sentiments against a share price and BSVM <GO> tracks social media velocity among a portfolio of stocks.
As the amount of data increases exponentially, the future of alternative trading is looking increasingly bright. Firms that are starting to embrace this are cutting through the clutter and uncovering new insights.
