What Is a Prediction Market? A Plain Explanation

A prediction market is a place where people buy and sell contracts that pay out based on whether something happens. That is the whole idea. The interesting part is what falls out of it: because the contracts trade, they have a price, and that price behaves like a probability.

Here is the simplest version. Somebody writes a question with a clear answer: Will the unemployment rate reported for October be 4.3% or higher? A contract is created that pays one dollar if the answer turns out to be yes, and nothing if it turns out to be no. People trade that contract. If it changes hands at 62 cents, the market is collectively saying the answer is about 62% likely to be yes.

That is the mechanism in one paragraph. Everything else is detail about how the question is written, how the payout is funded, and how the answer gets decided.

Why the price behaves like a probability

Think about what you are doing when you pay 62 cents for a contract that pays a dollar. You are risking 62 cents to win 38. For that to be a sensible trade, you have to believe the event happens more than 62% of the time. If you thought it was a coin flip, 62 cents would be a bad price and you would sell instead.

Now put thousands of people with different information and different opinions in the same place, all able to buy or sell. If the price drifts to 80 cents while most participants think the real chance is nearer 60%, selling becomes attractive and pushes it back down. The price settles where the weight of money is indifferent. That resting point is the market's estimate.

This is not magic, and it is not always right. It is an aggregation mechanism with a cost attached to being wrong, which is a meaningfully different thing from a survey where being wrong is free. More on that distinction in prediction markets versus polls.

The three things that make a market work

A question with one answer. The question has to be settleable. "Will the economy do well next year" is not a market. "Will US real GDP growth for 2027, as first reported by the BEA, exceed 2.0%" is, because there is a specific number published by a specific source on a specific date.

A funded payout. Somebody has to have the dollar. In practice that means both sides post the money up front. If you buy Yes at 62 cents and somebody sells it to you, their 38 cents and your 62 cents together make the dollar that will be paid to whichever side is right. Nobody is extending credit. This is why you will sometimes see people describe the two sides as fully collateralised.

A resolution source. When the event happens, something has to decide the answer. On a regulated exchange that is usually a named data source written into the contract. On an on-chain venue it is usually an oracle process with a dispute window. Either way, the rule is fixed before trading starts. Resolution deserves its own guide, because more disappointments come from resolution rules than from bad forecasts.

What you are actually buying

You are not betting with the house. There is no bookmaker setting a line and taking the other side of every trade. You are trading with other participants, and the venue takes a fee or a spread for running the place. That changes the economics: a sportsbook's margin is built into the odds it offers, whereas an exchange's cost is more visible and usually smaller. The comparison is worth doing carefully.

It also changes what the price means. A bookmaker's odds reflect its risk position as much as its opinion. An exchange price reflects where supply and demand balance among everyone trading.

What the price does not mean

A price of 62 cents does not mean the event will happen. It means that, weighted by money, the market thinks it is about 62% likely. If it does not happen, the market was not necessarily wrong. Events at 62% fail to happen 38% of the time, and a market that is right about that ratio over hundreds of questions is doing its job perfectly.

This is the single most common misreading, and it cuts both ways. People point at a market that said 15% and then saw the thing happen, and call it broken. Fifteen per cent things happen roughly one time in seven. If they never happened, the price would be wrong.

Judging whether a market is any good requires looking at many resolved questions at once and asking whether things priced at 30% happened about 30% of the time. That is calibration, and it is the only honest scoreboard.

Where the numbers on this site come from

This site tracks two venues continuously — Kalshi and Polymarket — normalises their prices into probabilities, and looks for the places where they disagree about the same real-world event. Nothing here is a recommendation, and nothing here is a forecast of our own: it is measurement of what those markets are currently saying, plus the arithmetic needed to compare them fairly.

If you want to see the mechanism in motion rather than described, the movers board shows which probabilities changed most in the last day and by how much. Start there, pick one market that interests you, and read its page. The abstraction becomes concrete quickly.

Where to go next

If you are new to this entirely, read how an event contract pays out next, then turning a price in cents into a probability. Those two cover ninety per cent of what you need to read any market page on any venue without guessing.