How prediction markets work

Prediction markets let people trade contracts that settle on real-world outcomes. How pricing, settlement and resolution work, and where the model breaks.

TL;DR

A prediction market trades contracts that pay a fixed amount if an event happens and nothing if it does not. The price therefore reads as a probability: a contract trading at 65 cents implies roughly a 65% chance in the market’s view. The hard part is not pricing but resolution — deciding what actually happened, and who decides.

Price as probability

A typical contract pays one dollar if the event occurs and zero if it does not. If it trades at 65 cents, buyers are paying 65 cents for a dollar that arrives only in one outcome, which implies a 65% probability.

This is the appeal of the format. Instead of a survey or a pundit, there is a number backed by money, updating continuously as information arrives.

The reading is not perfectly clean. Prices include the cost of capital tied up until resolution, and thin markets can sit away from fair value simply because nobody has bothered to correct them.

Where the counterparty comes from

Every contract needs someone on each side. Order book venues match buyers of yes against buyers of no directly, which is efficient where there is genuine two-sided interest.

Automated market makers quote both sides from a formula and a liquidity pool, guaranteeing a price even in markets nobody else is trading. That makes long-tail markets possible, at the cost of wider effective spreads and liquidity provider exposure.

Resolution is the hard problem

Pricing is straightforward compared with settlement. Someone has to decide what happened, and every design is a different answer to that question.

Some venues designate an authoritative source in advance — an official result, a named data provider. Some use a decentralised oracle where token holders vote and are penalised for voting against the eventual consensus. Some retain discretion for the operator.

Almost every serious dispute in prediction markets has been about resolution rather than price: an ambiguous question, an event that half-happened, or a source that never published. Reading the resolution criteria before trading matters more than reading the odds.

Regulated and permissionless venues

Two models coexist. Regulated event exchanges operate under a financial regulator, take fiat, restrict which markets they may list, and settle through traditional infrastructure.

Permissionless venues run on-chain with stablecoin collateral, list markets far more freely, and are accessible from anywhere with a wallet. They carry the same smart contract and oracle risks as other DeFi protocols, plus the resolution risk specific to this format.

The trade is familiar: more freedom in what can be listed, less recourse when something goes wrong.

Frequently Asked Questions

1.Are prediction market prices accurate forecasts?

They are competitive estimates, which historically beat pundits and often match or beat polls on well-traded questions. Accuracy depends on liquidity: a market with little volume is one person’s opinion with a price attached.

2.What happens if an event is ambiguous?

The resolution mechanism decides, and this is where disputes concentrate. Some venues have formal dispute periods and escalation; others resolve by operator decision.

3.Can I sell before the event resolves?

Yes, wherever there is a market. Most trading is exit before resolution rather than holding to settlement, which is why liquidity matters as much as the odds.

4.Why do the same odds differ between platforms?

Different user bases, different liquidity, and sometimes subtly different question wording. Those gaps are what cross-platform arbitrage tries to capture, and the wording differences are why it is riskier than it looks.

See this in the data

Related explainers

What Is a Funding Rate?How to Read Perp DEX Volume