Extracting Mid-Frequency Alpha from Analyst Ratings Sentiment

The RavenPack Wall Street Sentiment captures analysts’ sentiment, by leveraging their recommendations through real-time textual data and structured corporate events.

By targeting upgrades and downgrades from analyst ratings, the strategy leverages the increased investor attention and buying/selling pressure, capturing the tone and sentiment of the recommendations.

Using RavenPack’s proprietary NLP technology, billions of news are normalized and scored for sentiment, relevance, novelty, and factuality. These signals are then filtered and consolidated, and finally aggregated at the company level.

Actual and simulated performance of the Wall Street Sentiment portfolios for daily, weekly, and monthly time decay lengths from July 2007 to May 15, 2025. Portfolios are constructed for each of the six trading universes divided by region and market capitalization.

Backtested Performance (2007–2025):

The approach is applied across six regional and market cap universes, including U.S., Europe, and Asia-Pacific. In backtests spanning nearly two decades, the strategy consistently delivered strong returns with low portfolio turnover.

For Mid and Large Cap stocks:

• US: 3.3% annualized return | IR: 0.67 | Effective holding period: 5.4 days
• Europe: 8.1% annualized return | IR: 2.09 | Effective holding period: 5.5 days
• Asia-Pacific: 28.4% annualized return | IR: 4.96 | Effective holding period: 5.8 days

Why it works

The signal captures how market attention and sentiment timing influence price discovery around analyst actions. It benefits from underreaction to novel, credible information and avoids overfitting by aggregating at the entity level with pre-defined decay functions.

Download the full whitepaper to dive deeper into the model architecture and results.