How Retail Sentiment Actually Predicts Small-Cap Moves (The Data)
Retail sentiment as a stock-price predictor is either dismissed as noise or over-hyped as a leading indicator. The truth is more nuanced. Depending on the specific sentiment signal, the small-cap sub-universe, and the market regime, retail sentiment ranges from genuinely informative to actively misleading. Here's what the empirical work actually shows.
Three different sentiment signals
Not all sentiment is the same. Three distinct signals get lumped together but behave very differently in practice: (1) mention velocity (the change in mentions of a ticker per unit time), (2) author breadth (the number of unique authors mentioning the ticker), and (3) polarity (the ratio of positive to negative language in mentions). Studies find velocity leads price moves in small-caps; polarity often lags; breadth is the middle ground.
Mention velocity has the strongest leading signal
The most robust finding across peer-reviewed sentiment studies (Da-Engelberg-Gao 2011, Chen-De-Hu-Hwang 2014, and many follow-ups) is that changes in retail attention lead small-cap price moves by 1-5 sessions on average. Not the level of attention — the change. A stock going from 10 mentions/day to 100 mentions/day is a bigger signal than a stock steadily at 200 mentions/day. This is why StockRiot's Riot Score weights velocity heavily.
The market cap threshold matters
Retail sentiment predicts price moves most strongly in the sub-$500M market cap band. Above $2B, institutional flows dominate and retail sentiment becomes noise. The intuition: in liquid mid- and large-caps, retail volume is a small fraction of total volume, so retail behavior doesn't move prices. In small-caps, retail can be 30-70% of daily volume, which is enough to actually change price.
When sentiment misleads
Sentiment fails as a predictor in two specific patterns: (1) sentiment chasing price — mentions spike AFTER the price has moved significantly, meaning the sentiment is reactive not predictive; and (2) sustained high-level sentiment without acceleration — a stock with consistently 500 mentions/day but no meaningful velocity change is priced-in retail attention that carries no fresh information.
How the Riot Score handles these
The Riot Score is a weighted composite: 50% velocity, 30% breadth, 20% polarity. Scores are computed on a 0-100 scale with velocity contributing most because that's where the empirical predictive power sits. See the methodology page for the full formula.
Get StockRiot updates by email
One email. No spam. Unsubscribe anytime.
We use your email to send editorial updates. See our privacy page.Frequently asked questions
Does retail sentiment always predict price?
No. It's a signal, not a prediction system. It's informative in specific conditions (small-caps, velocity-driven, not chasing price) and noise in others.
How much lead time does retail sentiment give?
1-5 sessions on average in the sub-$500M cap band, per academic studies. Some individual events have longer lead times; some have essentially zero. It's a distribution, not a fixed number.
Should I trade purely on sentiment signals?
No. Sentiment is one signal in a stack. The stronger setups combine sentiment velocity with technical breakout patterns, catalyst news, or short-interest data.
How is Riot Score different from just counting mentions?
Raw mention counts don't distinguish velocity (change) from level (stable state). Riot Score weights velocity heavily and adjusts for breadth (unique authors) and polarity. It's meant to isolate the specific dimensions of sentiment that historically lead price.
Where does StockRiot pull sentiment data from?
Public forums where retail traders congregate: Reddit finance subs, StockTwits, Discord open channels, and public Twitter/X. Only publicly-accessible sources; no scraping of private groups or paid Discord servers.