A prediction market is an exchange-traded market for the outcomes of future events. Traders buy and sell contracts whose payoff depends on whether a specific event occurs — a candidate winning an election, an economic indicator crossing a threshold, a sports team lifting a trophy, or a product launching by a specified date. Because prices reflect the aggregate probability estimates of everyone trading, prediction market prices are widely used as consensus-based forecasts of the underlying events.
How does a prediction market work?
Each market is defined by an event with a resolvable outcome. The simplest and most common form is a binary market with a YES contract and a NO contract on a question such as "Will the US CPI print above 3.0% for October 2026?". Traders take positions by buying the contract on the outcome they expect. The market price of each contract sits between 0 and 1 (or 0% and 100%) and can be read as the implied probability of that outcome. When the event resolves, the winning contract pays its face value; losing contracts pay zero.
Multi-outcome markets extend the same design: for a market with N mutually exclusive outcomes, the sum of the N contract prices equals the face value (1.00 in most designs), preserving the interpretation of each price as a probability. Continuous markets (scalar markets) let contract payoffs vary linearly with a reported real-world value — useful for forecasting indicators rather than binary events.
How are prediction market prices interpreted as probabilities?
If a YES contract on "Will X happen?" trades at $0.63, the market is currently pricing the probability of X at about 63%. This reading is only valid when three conditions hold:
- Contract payoffs are binary — either 0 or the face value, no in-between.
- The denomination is a stable unit (fiat or a stablecoin), so nominal price movements track the underlying probability rather than currency effects.
- The market is liquid enough that quotes are not dominated by a single large participant.
Under those conditions the price is the market's implied probability. It is not a guarantee — many markets end up on the "wrong" side of what the price implied — but it is the best summary statistic of the collective view at that moment.
Are prediction markets accurate?
Peer-reviewed research going back to the 1980s finds prediction markets are at least as accurate as expert forecasts and often more accurate, especially over medium-term horizons of days to months. The Iowa Electronic Markets, one of the earliest academic prediction markets, matched or beat traditional election polls in most US presidential races between 1988 and 2012.
Accuracy scales with liquidity, incentive alignment, and question precision. Thinly traded markets and ambiguously worded questions produce noisier forecasts. Well-designed markets on questions with objective resolution criteria — court rulings, government data releases, sports outcomes — tend to produce sharp probability estimates that update quickly to new information.
How are prediction markets resolved?
Each market specifies a resolution source when it opens. Common sources include:
- Official government releases (BLS, Fed, PBoC, BOJ, and other central-bank or statistical-agency publications)
- Primary news reports from named outlets (Reuters, Bloomberg, AP, Xinhua, Nikkei)
- On-chain data (block explorers, DEX aggregators, oracles for token prices)
- Court filings, election commission announcements, sports league official results
- Independent oracle networks (Chainlink, UMA, Pyth, and others)
Well-designed platforms publish the resolution proof — a link, a hash, a screenshot, a transaction — so any user can independently verify the outcome. Contested resolutions are routed through a formal dispute window before payouts are finalized.
For deeper reading, see the companion guide on prediction market oracles.
What is a prediction market oracle?
An oracle is the mechanism a prediction market uses to determine and record the outcome of an event. Oracle designs sit on a spectrum:
- Authoritative oracle. A single trusted operator posts the outcome. Fast and simple, but the operator becomes a central point of failure and censorship.
- Optimistic oracle. Anyone can propose an outcome; other stakers can challenge within a dispute window. Only unchallenged outcomes settle without arbitration. Slow but manipulation-resistant.
- Hybrid oracle. A live data feed proposes an outcome, an operator or verifier confirms it, and users can raise a dispute. Combines speed of the live feed with the check of a human verifier.
What is the difference between a prediction market and a betting exchange?
A betting exchange is a subtype of prediction market focused on sports outcomes with the primary purpose of gambling entertainment. Prediction markets more broadly cover political, macroeconomic, scientific, and corporate events, and their primary purpose is information aggregation — surfacing the crowd's best estimate for use in research, journalism, corporate hedging, and policy analysis. In many jurisdictions the two are regulated under different frameworks. See the companion guide on prediction market vs. sportsbook.
Who uses prediction markets?
Users typically fall into five overlapping groups:
- Quantitative traders who profit from mis-pricings and provide liquidity in the process.
- Journalists and researchers sourcing crowd-based probability estimates for stories and papers.
- Corporate risk managers hedging exposure to elections, regulatory decisions, and macro events.
- Professional forecasters using market prices to benchmark their own accuracy.
- Retail participants engaging for information and, on some platforms, entertainment.
Are prediction markets legal?
Legality varies by jurisdiction and by the type of contract. In the United States, event contracts on registered Designated Contract Markets are permitted under CFTC oversight. In the United Kingdom, spread betting and Betfair-style exchanges operate under Gambling Commission licensing. In many Asian jurisdictions decentralized prediction markets operate under general financial or gambling frameworks; some regulators — notably Singapore's Gambling Regulatory Authority in January 2025 — have explicitly restricted access to non-domestic platforms. Users are responsible for verifying compliance in their own jurisdiction.
The country-by-country picture in Asia is covered in the companion guide on prediction market regulation in Asia.
A brief history of prediction markets
Prediction markets are older than they look. Organized betting on political outcomes was a mainstream activity in the United States and United Kingdom from the 1860s through the 1930s, with the "election betting" section of the New York Times often larger than the sports section. The modern era of academically studied prediction markets begins with the Iowa Electronic Markets in 1988, followed by corporate-internal markets at Hewlett-Packard and Google in the early 2000s. The 2010s introduced blockchain-based platforms (Augur, Gnosis, Polymarket, Kalshi), broadening market coverage from politics to a wide range of macro, sports, and crypto-native events. The 2020s have seen prediction markets emerge as a serious data source for major news organizations.