At Context Analytics, we previously demonstrated that combining sentiment from X (formerly Twitter), StockTwits, and News at a daily cadence can produce robust trading signals. In this follow-up, we extend our methodology to the monthly level, constructing longer-hold portfolios based on persistent sentiment trends.
From Daily Chatter to Monthly Signal
To move beyond short-term conversation, we developed custom monthly sentiment scores for each source. For social data, our s-factor feed can be aggregated to generate long-term signals—particularly from raw metrics like Raw-S and S-Volume. The custom scores are as follows:
- Monthly Z-Scores (Twitter & StockTwits)
We calculate a z-score based on the 30-day sum of Raw-S, standardized over a trailing 90-day window, with days of zero volume treated as zeros:
- Monthly News Positivity Ratio
For each stock, we count the total number of positive phrases in news articles over a 30-day window—treating days with no news as zero—and divide that by the total number of phrases for the same period to calculate a positivity ratio. This ratio has consistently shown strong predictive power for both long and short signals.
Portfolio Construction and Methodology
We sort stocks into monthly quartiles by:
- Monthly Twitter Z-score
- Monthly StockTwits Z-score
- Monthly News Positivity Ratio
Then construct:
- Long and short portfolios for each individual data source (top/bottom quartile)
- Consensus long and short portfolios: Stocks ranked in the top (or bottom) quartile across all three
- One theoretical long/short: Portfolio with long on Consensus Top and short on Consensus Bottom
Rebalancing occurs on the last trading day of each month, holding positions until the following month-end.
Signal Correlation
➤ Signals Are Uncorrelated
Our correlation heatmap reveals very low correlation between news sentiment and social platforms, while Twitter and StockTwits show slightly higher correlation at the monthly level compared to daily—but still not enough to produce overlapping portfolios. This low cross-source correlation supports effective diversification and justifies combining signals.
Portfolio Performance
➤ Positive Consensus = Signal Amplification
Our consensus top quartile long only portfolio —constructed when all three sentiment signals align positively—outperformed SPY by 29% cumulatively since August 2021. It also delivered stronger returns with superior risk-adjusted metrics.
➤ News Outperforms Social
Among individual sources, news sentiment delivered the widest performance spread between top and bottom quartiles. While Twitter and StockTwits also provided meaningful signals, the consistency and depth of news-based sentiment stood out—underscoring the value of persistent media tone in return prediction. Additionally, we see the top news quartile outperformed the SPY on its own.
➤ Robust Portfolio Size
The top consensus quartile averaged 22 stocks per month, ensuring a well-diversified and actionable portfolio size.
Final Thoughts
Our monthly sentiment strategy reinforces a key principle at Context Analytics, the leader in unstructured financial data: combining diverse data sources enhances alpha generation over both the short and long term. News sentiment provides robust standalone signals, and when integrated with social media data, we observe compounding effects that strengthen performance and reduce signal redundancy.
We continue to expand this research across sectors, timeframes, and market environments. For more information, visit Context Analytics and explore how our tools can support your investment process.
