Stocktwits S-Score Close-to-Close Quintiles

Context Analytics (CA) has ingested messages from Stocktwits, a social media platform for discussing financial securities, for over 10 years. CA scores and aggregates sentiment by security from each message over a rolling 24-hour window and creates S-Factors, which describe the conversation quantitatively. The S-Factor feed is one of Context Analytics’ longest running products and includes the S-Score, which is a description of sentiment at the security level.

S-Score is calculated by security and gives a cross-sectional view of Social Sentiment without bias towards message volume because the security’s sentiment is compared to its historical baseline.

Securities with extreme positive sentiment are expected to outperform the market, while securities with extremely negative sentiment are expected to underperform.

The charts below uses the S-Score from 20-minutes prior to market close (3:40pm ET) to group the securities into quintiles. Securities within each quintile are equally weighted and the Q5-Q1 is an equally weighted Long/Short between the top (most positive) and bottom (most negative) quintile.

The quintile plot above demonstrates a pattern between the S-Score and the subsequent daily close-to-close return period. When the S-Score is high, meaning sentiment from users on Stocktwits is more positive than normal, then the security outperforms. Conversely, when the S-Score is low, it underperforms. There is a significant spread between Quintile 5 and Quintile 1 which generates a stable Long/Short strategy with a Sharpe ratio of 1.64. Based on the average count of each quintile, the graph above includes the returns of over 1900 securities each day.

Using Message Volume as a Filter

Many securities have only one or two messages each day. This could be the opinion of one or two users which can skew sentiment values. For the next graph we only include securities that have at least three messages (S-Volume >= 3). The plot below is the same strategy as the plot above, however it removes the low message volume securities.

With the message volume filter applied, the spread between Quintiles expanded, offering stronger returns on both the Long and Short side even with the number of securities per quintile decreasing by over 40%. The Long/Short (Q5-Q1) portfolio outperformed the Long/Short without the volume filter by 27% cumulatively over this 3.5 year period.

Context Analytics’ S-Factor feed is the company’s most mature product. SV-Score, another metric in the S-Factor feed describing abnormal message volume, can be used as an alternative to S-Volume. CA’s offering of 10+ years of out of sample data and a suite of sentiment factors makes this dataset unique and ideal for back testing. With this dataset, users can harness the power of social media in financial markets. For more information visit www.contextanalytics-ai.com or reach out to us at ContactUs@ContextAnalytics-AI.com