Glossary
Prediction market
A prediction market is a market where people trade contracts on a future event. Each pays a fixed sum, usually $1, if the event happens and nothing if not, so the price is read as the crowd's estimate of its probability.
Also called Event contract, Event market, Information market, Forecasting market, YES and NO shares
See also
A prediction market is a market where people trade contracts on the outcome of a future event: an election, a data release, a match, the price of bitcoin at a set hour. Each contract pays a fixed sum, usually $1, if the event happens and nothing if it does not, so the price reads as the crowd’s estimate of a probability.
The academic reference is the Iowa Electronic Markets, a real-money market the University of Iowa runs for research and teaching. The best-known venues today are Kalshi, a US exchange, and Polymarket, the crypto-based platform.
How a prediction market works, from question to payout
- A market asks a yes-or-no question with written rules.
- Two shares exist, YES and NO. Bought together they cost $1 before spreads and fees, because exactly one will pay.
- A YES price of $0.62 implies a 62% probability, though it stays a market price, errors included.
- The outcome is settled. On Polymarket each winning share then pays $1 and losing shares become worthless.
In an order book, bids meet asks. Polymarket runs a hybrid: an operator matches orders off-chain, then the trade settles on-chain on Polygon. In an automated market maker (AMM), as on decentralized exchanges, a formula sets the price from the quantities already sold. The classic version is the market scoring rule, from the economist Robin Hanson’s 2003 paper in Information Systems Frontiers.
Who decides the outcome: a token vote or a rulebook
Polymarket’s standard route is UMA’s optimistic oracle. Someone proposes a result and posts a $750 bond, lost if the proposal fails. Anyone can object during the next two hours. If nobody does, the result stands and the proposer recovers the bond plus a reward.
A dispute can end in a token-holder vote, and UMA’s documentation requires at least 65% of staked UMA to back one outcome.
Kalshi has no token vote. The CFTC designated KalshiEX as a contract market on 3 November 2020, and the exchange settles its contracts itself against a source agency. A KalshiEX filing of 13 October 2026 sends some edge cases to its Outcome Review Committee.
A worked example: what a 62-cent price means
This example is hypothetical and ignores fees. YES on event E trades at $0.62. Buying 100 YES shares costs 100 × $0.62 = $62.
If E happens, the shares pay 100 × $1 = $100, a gain of $100 − $62 = $38, or 61.3% on the stake (38 ÷ 62). If it does not, you lose the $62.
The spread adds a cost. With a best bid of $0.61 and a best ask of $0.63, the midpoint is $0.62, and buying at the ask makes the same 100 shares cost $63.
An AMM charges more as the order grows. Under Hanson’s rule with a liquidity parameter of $100, a new market starts with YES at $0.50. Buying 10 YES costs $5.12, an average of $0.512 per share. Buying 100 YES costs $62.01, an average of $0.620, and lifts YES to $0.731.
Real cases: heavy volume, a disputed vote, a drained front end
Size depends on who counts. Stifel put regulated US volume at $77 billion for September 2026 but estimated taker volume, its closer match to sportsbook handle, at about $17.7 billion, Casino.org reported. Macquarie’s total for the same month was $20.1 billion.
In March 2025 a Polymarket market of about $7 million, on a Ukraine minerals deal with Donald Trump, resolved YES though Ukraine had not officially agreed. Users alleged, without proof, that a large UMA holder had swung the vote. Polymarket said the market had not failed and, The Block reported, refused refunds.
On 29 October 2026, offshore Polymarket lost about $2.65 million to a compromised third-party script, after about $3 million in June. It pledged to reimburse users.
The risks of a prediction market go beyond a wrong call
A thin market is easy to move, and ambiguous wording can turn a settled-looking question into a fight over the rules. France’s gambling regulator warns of compulsive play, the illusion of competence and bettors able to influence the result. The CFTC says it has sued a US service member over contracts on the ouster of Nicolás Maduro and a Google employee over Year in Search contracts, both accused of trading on nonpublic information.
This page explains a mechanism and gives no betting advice.
Prediction market rules in the US, France and Italy, as of October 2026
In a complaint filed on 26 October 2026, the CFTC argues that event contracts are swaps under its exclusive jurisdiction and that federal law overrides state gambling rules. That is a party’s claim, and the lawsuit against Utah is open. Utah filed criminal charges against Kalshi, its chief executive and its chief operating officer on 28 October; those are accusations, which Kalshi calls an overreach.
France’s gambling regulator, the ANJ, treats these platforms as illegal gambling sites and on 16 July 2026 ordered internet providers to block polymarket.com. Italy’s customs and monopolies agency, ADM, lists polymarket.com and kalshi.com among blocked gambling sites, in its list updated 13 October 2026.
Prediction markets: what readers ask
Are prediction markets gambling?
It depends on the jurisdiction. The CFTC calls event contracts federally regulated swaps, Utah treats them as gambling, and France’s ANJ calls the platforms illegal gambling sites. As of October 2026 the US dispute is before several courts, with no final answer.
Are prediction markets legal in the US?
Federally, yes in the CFTC’s view: Kalshi is a designated contract market and the agency claims sole oversight. Several states disagree, and the CFTC says it has started proceedings against nine states besides Utah. The courts had not resolved the conflict by October 2026.
Are prediction markets accurate?
Often, with limits. Justin Wolfers and Eric Zitzewitz reported in the Journal of Economic Perspectives in 2004 that market forecasts are generally fairly accurate and usually beat moderately sophisticated benchmarks. That finding guarantees nothing for a thin market or a 15-minute contract.
How much can you lose on a prediction market?
On a simple share purchase, you can lose the whole stake and no more: $62 on 100 shares bought at $0.62. Fees and spreads add to that cost. On crypto-based venues, security failures and disputed resolutions are separate risks for users.