Social Sentiment (3H) score

Social media sentiment from the last 3 hours. Tracks very recent retail investor mood and viral content on Reddit and Twitter.

How StocksScores scores it

Only rows from the last 3 hours with at least one mention count. Each row gets a 0-100 reading: 60% from the overall score mapped as 50 + score x 40, 30% from the positive share of positive+negative mentions, 10% from the balance of positive vs negative intensity (50 +/- 20). Thin rows are pulled toward 50 (kept at 40% strength under 3 mentions, 70% under 10, 85% under 20), and extreme readings off few mentions are damped. Rows are averaged with weights of log(1 + mentions) times an exponential recency decay with a 1.5-hour time constant. The distance of that average from 50 is then scaled by total positive+negative mentions in the window: 70% under 10 mentions, rising to 80% (10+), 85% (50+), 90% (100+), 95% (200+) and 100% at 500+, so heavy one-sided chatter moves the score further than light chatter. The windows are nested (3 hours inside 1 day inside 3 days inside 1 week), so a recent item counts in every window it falls in; this is a deliberate recency tilt.

Data: Finnhub /stock/social-sentiment for the last 7 days: per-period rows with mention count, positive and negative mention counts, positive and negative scores, and an overall score. When Finnhub leaves the overall score at 0 but the row has a signal, the server derives it as (positive - negative) / (positive + negative) mentions.

Not scored when: No Finnhub social data, or no row in the window with a mention count above zero.

Weight in the stock score

News sentiment is shown beside every score and never counted in the composite: in our own test it added nothing to the ranking. This scorer carries 5% of the news reading.

The full method — timeframes, the sector ranking and how missing data is handled — is on how the stock score works.

No US stock worth $1B or more has a Social Sentiment (3H) reading in the latest nightly batch.

Related scorers