At a very well-attended New York event, RavenPack recently launched a new self-service data and visualization platform - allowing us to move beyond the quantitative community with our data offerings. If you missed this event, you’ll get another chance to see the new self-service data and visualization platform in action at our upcoming London Event.
As part of our internal research, we recently took the new dataset for a spin - looking at one of the new key features - Event Relevance. Event Relevance provides information on the prominence of an event within a document and goes beyond our previous Relevance analytic that focused purely on the importance of a detected entity. As part of our analysis, we also considered the significance of event novelty by consulting our Event Similarity Day score.
As shown below, we generally find that high event relevance events (with a 90 to 100 score) significantly outperform low relevance events. However, the latter seem more predictive for small cap vs. large cap companies - possibly driven by the fact that it’s more difficult for smaller companies to make headlines or to be the focus of a news story than it is for larger companies.
We also reconfirm that more novel events provide stronger risk-adjusted performance across our four size & region universes. However, this comes with the tradeoff of lower event volume as we filter more aggressively on the number of days required since a similar event was detected in the past (up to 365 days).
Overall, results show RavenPack's new data set yields higher risk-adjusted returns in both the U.S. and Europe across both large/mid-cap and small-cap companies. Specifically, we find:
- The observed increase in event volume leads to more signals and larger portfolios
- The new RavenPack Analytics data generates higher risk-adjusted returns across region and size with an increase in Information Ratio of up to 50%. These results are achieved by filtering for novel and highly relevant events within a story
- The sentiment indicators are statistically significant at the 95% level across region and size, demonstrating both their robustness and ability to generate alpha.
Request the full white paper titled “Introducing RavenPack Analytics for Equities” for additional details about our methodology.