Where a market price comes from
Ask who sets the price on a prediction market and the answer is strange: nobody. On a question quoted at 68, there is no analyst who chose 68, no committee, no author at all. Sixty-eight is where two groups of people who disagree about the future currently balance.
That balance point is the product. The rest of this piece is how it gets made.
A number nobody set
Most public numbers about the future have an author — a forecaster, a desk, a model, a methodology page. A market price is the exception. It is left behind by trading, the way a water level is left behind by water.
That sounds like a weakness. It is the entire point.
The book
A prediction market trades a simple contract: it settles at 100 if the event happens and at 0 if it does not. The price in between is read as a probability — 68 means the market currently treats the event as a 68% chance (Wolfers and Zitzewitz 2004).
The venue keeps an order book with two sides. People who think the event is more likely than the current price buy Yes. People who think it is less likely buy No, the mirror position. Anyone can post a standing offer at a price of their choosing; that participant is a maker. Anyone can accept an offer already resting on the book; that participant is a taker. Every trade needs one of each.
To buy Yes at 68, you have to believe the true chance is above 68. Whoever fills you believes it is below. A trade is a disagreement made explicit — priced, sized, and time-stamped. The book is that disagreement, stacked: conviction on one side stacked against conviction on the other, at every price.
The gap between the highest price a buyer will pay and the lowest price a seller will accept is the spread, and the number you see quoted lives inside it.
Not every venue keeps a standing book. Some quote through a published pricing rule instead — an automated maker, run in the open, that moves its posted price as people trade against it. The plumbing differs; the authorship does not. On either design, nothing but trading moves the number.
Why the balance point is information
The reason this number is worth anything: every move in it was paid for.
Suppose the market quotes 68 and you know something that makes 80 the better estimate. Talking does not move the number. Trading does. Each trade moves the price exactly as far as someone was willing to back the disagreement.
That bound is the mechanism. A stated opinion is free, so opinions are abundant and carry no weight of their own. An order is not free — it is exposure to being wrong — so the book weighs every belief by what its holder will risk on it. The exposure that does the work is the commitment, not the currency: a dated, public number that will be graded either way — a weight venues where people trade purely for play impose just as fully.
The filter also pushes back against error. Whoever moves the number away from the evidence is, by doing so, offering better terms to everyone who disagrees; the further the price is pushed, the more attractive correcting it becomes. But be precise about what that guarantees. In one field experiment, researchers deliberately shocked prices with uninformed trades across 817 markets on a large play-money venue and watched what happened: prices reverted — quickly in the first week, more slowly after — and distortions were still detectable two months on. The markets with more traders, more volume and an outside reference point resisted best, and a follow-up on markets redeemable in dollars found the same pattern (Rasooly and Rozzi 2025, working paper). The pushback is real. Its speed is bought with depth.
Economists have formalized when this works: Wolfers and Zitzewitz (2006) show that under plausible conditions a market price closely tracks the central belief of the people trading it. Whether that number turns out to be accurate is a separate, empirical question with a large published record. That record — where these prices have been right and where they have not — is the subject of Are prediction markets accurate?.
What moves the price — and what can’t
What moves it: information in the hands of someone willing to trade on it. An injury report, a court filing, a data release — the moment it reaches a person with an account and conviction, it enters the book, and the number moves. Time can move it too, on questions with a deadline: every day an event fails to happen is itself information, and traders reprice it.
What can’t move it: attention without orders. A viral thread, a confident pundit, a surge of enthusiasm — if nobody trades, the number does not move. That is the useful difference between a market price and a trending topic. One can be pushed by noise alone. The other has to be bought.
Depth is not a cosmetic feature of a venue; it is part of what the number means. The tick data agree: a 2026 study of 30 billion order-book events on the largest on-chain venue found depth varies systematically with a market’s duration, price and volume — and that spreads widen on unlikely outcomes (Dubach 2026, preprint).
Read the number correctly
A market price is not a verdict, and it is not anyone’s opinion. It is the current balance of everyone willing to be scored on the question, recomputed every time someone disagrees enough to act — the level the trading leaves behind.
That is why the number deserves a different kind of attention than a take or a poll. Not because it is always right — it is not, and the record says so precisely — but because of what it costs to move.
So the next time you see a 68, don’t ask who set it. Ask who is willing to move it. That question is the whole difference between a number somebody wrote and a number somebody backed.
Sources
- Wolfers, J., Zitzewitz, E. (2004). “Prediction Markets.” Journal of Economic Perspectives 18(2): 107–126. aeaweb.org
- Rasooly, I., Rozzi, R. (2025). “How Manipulable Are Prediction Markets?” Working paper, arXiv:2503.03312. arxiv.org
- Wolfers, J., Zitzewitz, E. (2006). “Interpreting Prediction Market Prices as Probabilities.” NBER Working Paper 12200. nber.org
- Dubach, P. D. (2026). “The Anatomy of a Decentralized Prediction Market: Microstructure Evidence from the Polymarket Order Book.” Preprint, arXiv:2604.24366. arxiv.org