presence of a highly positive event can help reduce slippage and lower your transaction
costs.
Measuring the item’s novelty, or uniqueness, can be an important addition to
increasing one’s responsiveness to news items. An algorithm may resume automatically
if the news appears repetitious or await a human response if it appears to have new
information. This is especially important as more and more sources are included in the
system. One would not necessarily want to react on the 50th repetition of the item.
You can also determine the importance of the article by examining some of the other
metadata for clues. Stories about mergers and acquisitions or earnings guidance might
mandate review, while the third update to a story might be something to filter out and
ignore. Similarly, market commentary might be discounted altogether and broker
recommendations by Morgan Stanley’s analysts should be ignored when analyzing
Morgan Stanley, the company.
Combining the above factors along with a host of other metadata can help
significantly enhance the defensive use cases for machine-readable news.
So what about offensive strategies? How can one generate alpha using news?
For more offensive strategies, the same metadata can be used for more aggressive
positioning. The main conclusions from published research can each be refined when
incorporating some of the intelligence garnered from newly available metadata such as
that from Thomson Reuters News Analytics. Academic and other research on the
impact of news center around three main ideas
- News flow, or the rate at which news is published, can predict volume and volatility.
Greater news rates would be followed by higher trading volume and increased price
volatility. - Pricing movements accompanied by news tend to gain momentum in nature while
those with a lack of news tend to revert to their previous trends/ranges. - The market tends to overreact on high news volumes and underreact on low news
volumes.
Increasing the signal strength in these strategies can be accomplished through a number
of techniques. First, on news flow being able to predict increases in trading volume and
volatility: Studies have shown that more relevant news and more emotive news (highly
positive and highly negative in tone) can better predict volume and volatility increases.
In addition, negatively toned news has greater impact than does positively toned news.
Also, the type of item such as an alert vs. an update to a story can significantly increase
the ability to accurately predict volume and volatility spikes. When filtering out market
commentary and broker recommendations (as outlined above) and less important, less
influential, and biased sources, as well as highly repetitious items, one can further refine
the predictive signals in news flow.
Second, on pricing movements with accompanying news flow suggesting momentum:
Upward pricing movements with positive news (or downward movements with negative
news) from credible sources with the appropriate filters to exclude ‘‘irrelevant’’ news
could increase confidence in pricing momentum and reversal strategies.
Third, with the market likely overreacting when lots of news is published on a
company and underreacting on small amounts of news: One wouldn’t suggest shorting
Incorporating news into algorithmic trading strategies: Increasing the signal-to-noise ratio 309