The Riot Score

Last updated 11 September 2026

The Riot Score is a number from 0 to 100 describing how far a small-cap or micro-cap stock’s retail-investor chatter has moved from that stock’s own normal level today.

It is an attention measurement. A Riot Score of 95 says an unusual number of people are talking about a ticker, with unusual intensity, relative to how much they usually do. It says nothing whatsoever about whether the company is good, whether the stock is cheap, or whether buying it is a sensible idea. We are deliberate about that distinction, and the rest of this page explains exactly how the number is produced so you can judge it rather than trust it.

The covered universe

StockRiot scores US-listed small-cap and micro-cap equities. Concretely:

The cutoff is not arbitrary. Retail attention is only a meaningful variable where it is large relative to the float and the daily volume. Ten thousand people discussing a mega-cap changes nothing about its price; ten thousand people discussing a $70 million company with a four million share float is the entire order book. Scoring both on the same scale would produce a list where the interesting names never surface — which is the failure mode of every raw mention counter.

Liquidity floor. A name must clear a minimum average traded dollar volume to reach the board at all. This is deliberately a floor on dollars, not shares: a stock trading three million shares at $0.004 is not liquid, and attention on it is not actionable in either direction.

Why not just count mentions?

Because raw mention counts measure company size, not news.

Even inside our own universe, raw volume ranks by size. The most-mentioned names on retail forums are the handful of well-known small caps that are always being discussed, for the same reason the most-mentioned words in English are “the” and “and”. A list ranked by raw volume tells you those companies exist. It does not tell you that a $40 million industrial name went from six posts a day to four hundred overnight — which is the only thing on that list a trader could not already have guessed.

So the Riot Score is relative to each ticker’s own history. Every component below is measured as a deviation from that specific stock’s trailing 30-day baseline, not against the market as a whole.

The five components

The weighted components are combined and mapped onto a 0–100 scale. The mapping is calibrated against the trailing distribution of scores across the covered universe, so the bands below keep a consistent meaning instead of drifting with overall market noise.

Reading the bands

What gets excluded before scoring

Sentiment data is easy to contaminate, and a tracker that does not filter is just a megaphone for whoever posts most. Before anything is scored we drop:

• Duplicate and near-duplicate text — the same message posted across many venues counts once per author, not once per posting.
• Very new accounts — accounts below an age and activity threshold are excluded from the author-breadth count, because they are the cheapest thing in the world to manufacture.
• Single-author concentration — where one account contributes an outsized share of a ticker’s daily volume, that excess is capped rather than counted.
• Tickers outside the covered universe — anything above the $2B cap ceiling, and anything below the average-dollar-volume floor, never reaches the board.
• Ambiguous cashtags — symbols that collide with common words are handled by context matching, and dropped where context is insufficient.

What the Riot Score deliberately does not include

Just as important as what goes in:

• No fundamentals. Revenue, margins, cash runway and dilution risk are not inputs. The score is not a quality rating and must not be read as one.
• No price targets, no ratings, no recommendations. We do not produce them.
• No paid placement. Position on the board cannot be bought. See Disclosures for how compensation is handled where it exists.
• No private or paywalled sources. Only publicly visible posts are counted. We do not scrape private servers, DMs, or subscription rooms.

Known limitations

Every sentiment methodology has failure modes. Here are ours, stated up front:

Attention is reflexive

Publishing an attention score can itself attract attention. We mitigate this by publishing measurement rather than selection — a board of what already happened, not a watchlist of what to do next — but the effect cannot be eliminated and you should assume it exists.

Sentiment scoring misreads sarcasm

Retail forums run on irony. Automated polarity scoring handles plain language well and sarcasm badly, which is one of several reasons polarity carries only a 15% weight rather than driving the score.

Coordinated promotion can still get through

Filters raise the cost of manufacturing a trend; they do not make it impossible. A well-resourced campaign across many aged accounts will register as real attention, because by any measurable definition it is real attention. That is precisely why we publish the lead/lag component — manufactured attention tends to arrive without a corresponding catalyst, and it shows.

Baselines break on corporate actions

Reverse splits, ticker changes and uplistings reset a name’s chatter profile. Scores in the days immediately following such an event are less reliable and are flagged on the board.

Cadence and corrections

The board is computed after the US regular session closes and published the same afternoon. Archived boards are dated and permanent — we do not silently revise or delete them.

When the pipeline gets something wrong — a mismatched ticker, a feed outage, a contaminated baseline — we publish a dated correction note on the affected page rather than editing it quietly. If you spot an error, tell us at editor@stockriot.com.

Get the daily board by email