How accurate are prediction markets? What the evidence shows
Prediction markets are often accurate on big, liquid questions and beat polls most of the time, but miss on long shots and thin markets. What research shows.
Read the articleA prediction market is an exchange where people trade on future events, and the price works as a probability. How it works, who runs them, why they matter.

A prediction market is an exchange where people buy and sell contracts on the outcome of future events, such as an election, a central bank decision or a product launch. Each contract pays out if the event happens. Because of that, its price works as a probability: a contract trading at 30 cents on a $1 payout means the market sees roughly a 30% chance.
In 2026 prediction markets went mainstream. Search interest in Kalshi alone rose by over 1,500% in a year (Rising Trends). This guide explains what they are, how they differ from betting and polls, and why forecasters take them seriously.
Take the question "Will the Democrats win the House in 2026?"
Every trade is a small bet on whether the price is too high or too low. Together, thousands of them pull the price toward what the best-informed traders believe. You can see this question, with live prices from two markets, on our 2026 midterm odds page.
For the full mechanics of order books, Yes and No contracts and settlement, see how prediction markets work.
A contract that pays $1 in one outcome and nothing in the other is worth, to a rational trader, roughly the probability of that outcome times $1. If a price drifts away from the true odds, someone who knows better can profit by trading it back. That incentive is what turns a market into a forecast.
In practice the price is an approximation, bent by fees, the gap between buy and sell prices, and a known tendency to overprice long shots (the favorite-longshot bias). We explain how to read it precisely in prediction market odds, explained.
| Type | Examples | Money | Notes |
|---|---|---|---|
| Regulated exchanges | Kalshi | Real money (dollars) | CFTC-regulated in the US |
| Crypto-based markets | Polymarket | Real money (stablecoins) | Global; US access through a regulated affiliate since 2025 |
| Play-money markets | Manifold | Play money | Anyone can create questions |
| Forecasting platforms | Metaculus, Good Judgment Open | No trading | Forecasters submit probabilities and are scored |
Only the first two are markets in the strict sense, with prices set by trading. Forecasting platforms produce probabilities without betting, and their best participants, the superforecasters, are the benchmark AI forecasters are measured against.
Often, and sometimes impressively, but not everywhere. Markets do well on liquid, high-profile questions, less well on thin markets and long shots, and they can be confidently wrong, as on Brexit in 2016. The evidence is in how accurate are prediction markets?
Sikt Intelligence is building an AI superforecaster: AI that researches a question and gives its own calibrated probability. We show it next to the market price, because when the two disagree, the gap is the signal. Explore live market odds on the biggest events on our odds page. Nothing here is financial advice.
A place where people trade on whether future events will happen. A contract pays $1 if the event happens, so its price in cents is roughly the market's probability in percent.
Legally and structurally it is different: in the US, regulated prediction markets are overseen by the CFTC as exchanges for event contracts, and traders trade against each other rather than against a bookmaker. Whether some contracts, especially on sports, count as gambling under state law is being decided in the courts in 2026.
Kalshi and Polymarket are the largest in 2026. Kalshi is a CFTC-regulated US exchange; Polymarket is a global, crypto-based market with a regulated US affiliate.