Play free
← All posts

How do prediction markets work? (one question, start to finish)

Every market anyone has ever opened began the same way: somebody wrote a question down. So the clearest way to see how prediction markets work is to follow one question through its whole life — from a sentence to a number to a fact.

(The category’s plain-English definition has its own page. This one is about the machinery.)

Day one: a sentence with a date on it

Say the question is: will the album drop by Friday? To become a market, that sentence needs two more things — a date it must answer by, and a named source that will do the answering. Not vibes, not a mod’s judgment call later: the ticket says, on day one, what counts as Yes and who says so.

That’s the whole entry fee for a question. If it can’t settle cleanly, it never opens.

The number appears

Calls arrive. Somebody who follows the artist takes Yes; somebody who’s watched three delays this year takes No. Every call presses the number toward that caller’s side, and it comes to rest where the pressing balances — say 64¢ on Yes. (Illustrative numbers, here and below.)

Nobody typed that 64 in. It’s the room’s combined read: probably, but no promises — and both halves of that sentence are visible at once, because Yes and No always total 100. One number carries the confidence and the doubt together.

News lands, and the number walks

Wednesday night, the artist posts something cryptic and Friday-shaped. Here’s the part worth watching: no one updates the price. The people update. Whoever reads the post first, and best, presses first — and the number walks from 64 to 71 while the rest of the room is still deciding what the post meant.

That’s the entire trick of the category. The price isn’t edited like a headline; it’s pushed, call by call, by people who each think they’ve read the moment a shade better than the room has.

Friday: the question stops being a question

The date arrives. The named source answers — the album is out, or it isn’t — and the market settles. The settling itself has its own quiet choreography, but the visible part is simple: the arguing ends, and right and wrong become facts with names attached.

And then the price, which worked so hard all week, dies. That’s fine. It did its job: it was the room’s honest read, updated live, for exactly as long as the future was still open.

What survives the question

The record. Every person who called it now has one more entry — right or wrong, at what confidence, on the record for good.

That’s the part majjha keeps and the part money never touched. Here the whole life above runs on Dots — free, no card, no money in or out — so the only thing a question leaves behind is proof of how well you read it. String enough of those together and you have the thing this whole machine was quietly built to measure.

Somewhere on the board right now, a question is mid-life, its number resting where the room last left it.

Go read one — and if you’d move it, move it.

Quick answers

How does a prediction market actually work?

A real question with a checkable answer opens for calls. People who disagree take Yes or No, and a live price forms between them — the crowd's read of how likely it is. On the settle date a named source answers, the question closes, and everyone finds out who read it right.

Who decides how a question settles?

Not the crowd, and not the venue's opinion — the source the question named on day one. A well-written question states, up front, exactly what counts as Yes and who says so.

What moves a prediction market price?

Disagreement. Every call presses the number toward the caller's side, and when news lands, the people who read it first press first. The price is where all that pressing balances right now.

What happens if my call is wrong?

On majjha: your record says so, and nothing else happens — there's no money in it to lose. A miss on the record is the honest cost, and it's also how the next read gets sharper.

— majjha