What Is a Prediction Market? How Contract Prices Become Probabilities

A prediction market is an exchange where people trade contracts that pay a fixed amount if a stated event happens and nothing if it does not, so the price of the contract can be read as the market's probability that the event occurs.

This page covers what those contracts are, how to read a price as a probability, where the two main US venues differ, and the specific ways that reading can mislead you. It is descriptive throughout.

The contract is the whole idea

A prediction market contract names an event and a resolution date. "Will the Federal Reserve change rates at the September meeting?" is an event. The contract settles at $1 if the answer turns out to be yes and $0 if it turns out to be no.

Because the payoff is fixed at $1, the price carries a direct reading. A contract trading at $0.37 is a market of buyers and sellers agreeing that roughly 37 cents today is a fair exchange for a dollar paid only in the yes case. That is a 37% implied probability.

This is what separates a prediction market from an ordinary futures market. A crude oil future settles at whatever oil costs, which can be any number. An event contract settles at one of two numbers, so its price lives between them and can be read on a probability scale without any modelling.

Reading the price as a probability, and the four adjustments

The plain reading, price equals probability, is a good first approximation and a poor last one. Four things sit between the two.

Fees. Most venues take a cut, either on trades or on winnings. A contract at $0.37 with a fee on settlement is priced for a payout below $1, so the implied probability is not exactly 37%.

The spread. There is a price to buy and a price to sell, and they are not the same. A contract quoted 36 bid and 39 ask has no single price. Which side you read changes the probability by three points.

Time value. A dollar paid in December is worth less than a dollar today. On a contract resolving months out, the price is discounted for that wait, which pushes it slightly below the true probability.

Liquidity. A price with $500 behind it is a quote. A price with $3m behind it is a market. The number looks identical in both cases.

The two US venues resolve the same question differently

Kalshi and Polymarket are the two venues most US readers will meet, and they differ in structure, in fee schedule and in how a position is collateralised and settled. Both list contracts on overlapping questions, and the same question can trade at different prices on each at the same moment.

Some of that gap is mechanical: different fees, different collateral, different users. Some of it is a genuine disagreement about the answer. Telling those apart is the interesting part, and it is why comparing venues is more informative than reading either one alone. Kresmion tracks contracts on both venues and computes a combined reading for the smaller set of questions that are listed on both, which is the subset where a venue to venue comparison is meaningful at all.

The wording matters as much as the venue. Two contracts that sound like the same question often are not. "Will BTC touch $80,000 before December 31" and "Will BTC be above $80,000 on December 31" are different contracts with different fair prices, because the first pays on any single moment above the line and the second only on the final print. Comparing them as if they were the same question produces a fake divergence.

Where the probability reading breaks down

The favourite longshot bias. In many betting and event markets, contracts priced very low resolve yes less often than their price implies. Kresmion's own calibration work across resolved prediction markets found the effect concentrated in the extreme longshot tail rather than spread across the middle of the range, so a 3% contract deserves more scepticism than a 45% one. See the calibration study.

Resolution ambiguity. The contract pays on the resolution source, not on what actually happened in the world. If the source is a specific agency release and that release is delayed or revised, the contract follows the source.

Thin far tails. The extreme strikes on a ladder are where volume is thinnest and the pricing is least reliable. A ladder of contracts at rising thresholds will usually have a well traded middle and two noisy ends.

One venue is one opinion. A single price is a single market's view, subject to that market's users and its fee structure.

A worked example of the ladder

Venues often list a whole ladder on one underlying: a set of contracts at rising thresholds, each asking whether the price touches that level by a date. Read together, a ladder is a probability distribution rather than a single number.

The touch probabilities should fall as the threshold rises, because touching a higher level is harder than touching a lower one. When they do not fall in order, something is wrong with the quotes rather than with the world, and that inconsistency is itself readable. A ladder that stays near symmetric above and below the current price is a market that is not committing to a direction.

Key takeaways

Frequently asked questions

Is a prediction market price the same as a probability?

Close, but not identical. The price is what buyers and sellers will exchange for a fixed payout, which is a probability adjusted for fees, the spread between bid and ask, the time until resolution, and how much money stands behind the quote. On a liquid market resolving soon, the difference is small. On a thin market resolving in six months, it can be several points.

Are prediction markets better than polls or forecasts?

They are a different kind of measurement rather than a strictly better one. A poll samples opinion, a model applies assumptions, and a prediction market prices what people will stake money on. Each is wrong in its own way. The market has the advantage that participants pay for being wrong, and the disadvantage that a thin market can be moved cheaply.

Why do Kalshi and Polymarket show different prices for the same event?

Partly mechanics and partly disagreement. The venues have different fees, different collateral, and different user bases, which can hold a gap open even when both agree on the answer. On top of that sits genuine disagreement. The gap is only meaningful once you have confirmed the two contracts really ask the same question, including the resolution source and whether it settles on a touch or on a final level.

Can a prediction market price predict a market move?

No, and it is not built to. It states a probability for a defined event with a defined resolution date. It says nothing about the path in between, nothing about any other asset, and nothing about what a price will do. It is a reading of the odds on one question, and it is a snapshot that changes as the order book changes.

---

Source: Kresmion prediction market coverage, spanning Polymarket and Kalshi contracts, with order book capture and a combined reading where the same question is listed on both venues. Kresmion Research.