Docs / Scoring / Detection
📞 How a call is detected
Thousands of messages an hour. Only a fraction are real calls. This is how we tell the difference.
ℹ️ Note
In short: a message becomes a call when it names a token and commits to it. Chat, hype, reposts and questions are ignored — which is why the ranking is not a popularity contest.
What counts as a call
A token we can identifyA contract, or a ticker we can resolve to one specific token with confidence. Ambiguity is thrown away rather than guessed.A real position"Aping this" is a call. "What do we think of this?" is not. Intent has to be there.Said by a person, in a room we watchBot spam, forwards and link-dumps do not create calls.First timeReposting the same token later does not mint a second call. Your entry is where you actually called it.What gets thrown away
Hype with no token"THIS IS GOING TO RUN" with nothing attached scores nothing.Ambiguous tickersIf a symbol maps to several tokens and nothing settles it, we drop the message rather than guess and pollute a record.Recaps and screenshotsTalking about a move that already happened is not calling it.Copy-paste raidsIdentical messages across rooms are collapsed, not multiplied.A large share of typical group traffic never becomes a call at all. That is deliberate. Every message we let through has to survive being scored later, in public, against price.
Why we throw away so much
🧹
A record you can trust
One bad extraction becomes a permanent lie on someone's profile. Silence is cheaper than a wrong call.
🛡
Ungameable by volume
If posting more created more chances to rank, the leaderboard would belong to whoever spams hardest.
⚡
Fast enough to matter
Detection happens as the message lands, so an early call is timestamped early — not whenever we got round to it.
⚠️ Important
We do not publish the gates. The exact signals and cut-offs that decide what passes are kept private, because they are also the instructions for slipping fake calls past them.