Context Analytics is the leader in processing and structuring textual data for sentiment analysis. Our S-Factor feed contains 17 metrics derived from Twitter data, each designed to capture a different dimension of investor sentiment.
At the center of the S-Factor suite is the S-Score - a normalized indicator showing how today’s sentiment compares to the asset’s own history.
It is defined as:S-Score = Exponential 24-hour sentiment sum compared to its 20-day mean (SMean) and volatility (SVolatility)
Another metric is the Raw-Score which is calculated from the exact same tweets using a similar formula. The difference lies in when those tweets were posted.
The S-Score applies exponential time weighting to the 24-hour sentiment window. A tweet from 10 minutes ago carries significantly more influence than one from 23 hours ago. Raw-Score, by contrast, treats all messages within the window equally.
This distinction seems subtle. But in markets driven by real-time information flow, timing is everything.
This raises a key question:
Can we isolate the value of recency itself—and use it to generate alpha?
Defining Information Shock
To answer this, we created a simple derived metric:
Information Shock = S-Score – Raw-Score
This value captures excess sentiment—the degree to which recent information diverges from the full-day average.
Both metrics draw from the same underlying tweets. The only difference is the temporal weighting applied during aggregation.
Methodology
We ran standard quintile analysis using Information Shock as the ranking signal.
Step 1 — Quintile Formation
Each day, we rank the universe by Information Shock at 3:40pm ET and form five equal-weighted quintiles.
Step 2 — Return Measurement
We measure Close-to-Close returns for each quintile, then compute the Q5−Q1 spread.
The Results: Freshness Matters
The Q5−Q1 spread using Information Shock is substantial and statistically significant.
Securities with elevated recent sentiment—those where the news broke late in the window—outperformed. Securities where sentiment had decayed over the day lagged behind.
This is striking because both groups were exposed to the same volume of social media discussion. The only differentiator was when that discussion occurred.
In short: identical information, processed with different time weighting, produces meaningfully different forward returns.
Why This Works
Social media sentiment decays rapidly. A viral post at 10am may be fully priced in by noon. A late-breaking development at 3pm may still carry signal into the next session.
Information Shock isolates this recency premium by stripping out the level of sentiment and focusing purely on its temporal distribution.
High Information Shock suggests:
Low Information Shock suggests:
Equal-weighted sentiment aggregation misses this distinction entirely. Exponential weighting captures it—and Information Shock quantifies it directly.
Key Takeaway
Time is a first-order variable in social sentiment analysis.
Our findings demonstrate that the recency of information—not just its direction or magnitude—carries independent predictive power. Information Shock offers a clean way to isolate and exploit this effect.
This is one of many ways to manipulate S-Factor data for short-term alpha using Twitter. For more information, visit www.contextanalytics-ai.com .
TL;DR
Fresh News = Better Signal
Context Analytics tested whether the timing of sentiment—not just its level—predicts returns.
We defined Information Shock as S-Score minus S-Raw, isolating how recently sentiment emerged within the 24-hour window.
Results (2016–present):
Key insight: In social media data, when information arrives matters as much as what it says. Exponential time weighting isn't just a smoothing technique—it's an alpha source.
