Updated July 2026

What is a prediction market?

A prediction market is a place where you buy and sell contracts that pay out based on whether a real-world event actually happens. That one mechanic turns a question about the future into a live price, and the price behaves like a probability you can read off a screen. This is prediction markets explained from the ground up: what the contracts are, how a price in cents becomes a percentage, why these markets aggregate information better than most forecasts do, where you can trade them, and the honest limits that keep a careful person from treating the number as gospel.

The definition, in one sentence

A prediction market is a market where you trade contracts tied to a yes-or-no question about the future, and each contract settles at one dollar if the answer turns out to be yes and zero if it turns out to be no. Take a question like "Will the Federal Reserve cut interest rates at its next meeting?" The market lists a YES share and a NO share. If you think a cut is coming, you buy YES. If you think rates hold, you buy NO. When the meeting happens and the outcome is known, the market settles: the winning side is worth a dollar each, the losing side is worth nothing, and your profit or loss is the difference between what you paid and what the share is finally worth.

That is the whole machine. The event can be an election, an economic number, a court ruling, a sports result, or the box office of a film, as long as it has a clear, verifiable outcome and a fixed date to settle on. Because the payout is always exactly one dollar, the price you pay carries a precise meaning, and that meaning is the reason these markets are interesting.

How a price becomes a probability

Here is the idea that makes everything else click. A contract that costs 60 cents is the market saying the event has roughly a 60 percent chance of happening. The logic is direct: nobody rationally pays 60 cents for a claim on one dollar unless they believe they will collect that dollar about 60 percent of the time. Pay too much relative to the real odds and you lose over the long run; pay too little and someone else takes the trade before you. So the price the crowd settles on is, in effect, its collective estimate of the probability. A share at 8 cents implies an 8 percent chance. A share at 91 cents implies 91 percent.

The two sides fit together cleanly. Because holding one YES share and one NO share guarantees you exactly one dollar at settlement no matter which way the event goes, YES and NO always add up to about a dollar. If YES trades at 60 cents, NO trades near 40 cents. Whenever that pair drifts away from a dollar, traders buy the cheap combination or sell the expensive one and push it back. So you never have to memorize both numbers: read one price in cents, and you have read the market's probability in percent.

A worked example

Say a market asks whether a specific interest-rate cut happens next month, and YES is trading at 62 cents. The market is estimating a 62 percent chance. You buy 100 YES shares, which costs 62 dollars. Two things can happen. If the cut arrives, each share settles at a dollar, your 100 shares are worth 100 dollars, and you are ahead by 38. If rates hold, the shares expire worthless and you are down the 62 you paid. Notice the shape of it: you risked 62 to make 38, which is exactly the trade you should be willing to make only if you believe the true chance is higher than the 62 percent the market is charging.

The NO side works the same way from the mirror. NO is priced near 38 cents, so a buyer there risks 38 to make 62, a trade that makes sense only if they think the real probability of a cut is below 62 percent. Every prediction market is that pair of bets meeting in the middle, and the point where they meet is the price on the screen.

Why the price aggregates information and keeps updating

A poll asks people what they think. A prediction market asks people to put money behind what they think, and that changes the quality of the answer. Participants who have done real work, hold private information, or simply feel more certain can back their view with size, and their trades move the price more than idle opinions do. The result is a money-weighted average of everything the crowd collectively knows, which is why economists describe these markets as machines for aggregating dispersed information into a single number.

The number is also alive. A market only reprices when someone trades, so every move traces to a reason: a poll drops, a data release prints, a headline lands, and a liquid market can fold that information into its price within seconds. A poll cannot update itself between surveys, but a market updates continuously, all day, for as long as it is open. That combination, a crowd with skin in the game plus prices that move on real news, is the core of how prediction markets work and why their output is worth reading.

Where you can trade prediction markets

Three venues are worth knowing, and they occupy different corners of the same idea.

• Polymarket is the large on-chain venue. It runs on the Polygon blockchain and settles trades in the USDC stablecoin, and it is the biggest venue by volume. Its market for the winner of the 2024 US presidential election processed more than three billion dollars in trading, which made it the most-watched prediction market event to date. Being on-chain means the order flow is public and anyone can inspect it, which is part of why researchers and journalists cite its prices.
• Kalshi is the regulated US exchange. Kalshi is a federally regulated exchange, approved by the Commodity Futures Trading Commission as a designated contract market, that lets people in the United States trade event contracts on economics, politics, weather, and more. Its contracts work the same way, priced between one and 99 cents and settling at a dollar, but they sit inside a US regulatory framework rather than on a public chain.
• SmartX is an AI research and trading terminal. Rather than being a single market, SmartX is a terminal built to research prediction markets and trade them from one screen, with tools that rank wallets by track record and surface what informed traders are doing. It is a layer on top of the markets, aimed at people who want the analysis and the trading in the same place.

Availability and legality differ by venue and by where you live, so the right starting point is checking what is open to you before anything else.

How they differ from polls and from sportsbooks

People reach for two familiar comparisons, and both are instructive because of where they break down. A poll is a snapshot of stated opinion, sampled from a group at one moment and reported with a margin of error. A prediction market is a live, money-weighted probability that reprices on every trade. The poll tells you the shape of opinion right now; the market tells you the odds of the outcome, updated continuously and already digesting the polls themselves. We compare the two in depth in prediction markets versus polls, but the short version is that they answer different questions and the careful reader uses both.

A sportsbook looks closer, because you are also putting money on an outcome, but the structure is different. At a book, the house sets the odds and bakes its margin, the vig, directly into both sides of the line, and you are betting against the house. A prediction market is a peer-to-peer order book: the price is set by traders meeting each other, not by an operator, and the cost of trading lives in the bid-ask spread rather than in a house edge stamped onto the number. That is why a deep market's midpoint tends to be a cleaner read of the true probability than a bookmaker line, though it is never guaranteed to be right.

The track record, and the honest limits

Prediction markets have a genuinely strong forecasting record when the conditions are right. The most-cited evidence comes from the Iowa Electronic Markets, a long-running academic market: across 964 polls over the five US presidential elections from 1988 to 2004, the market was closer to the final result than the polls about 74 percent of the time, and its edge was largest far ahead of the vote, where single polls are noisiest. That is a real, measured advantage, not marketing.

It comes with limits you should take just as seriously.

• Thin markets are noisy. On a contract with little volume, the last trade might be stale or set by a handful of people, and the price is not a considered crowd opinion so much as an artifact of nobody trading. Deep markets are far better calibrated than thin ones.
• Prices can be moved. On a low-liquidity market, one large trade can shove a price from 30 to 60 cents and spawn headlines suggesting the crowd changed its mind, when only one wallet did. Telling a well-informed whale from someone trying to move the number is genuinely hard in the moment, and manipulation is a real risk on shallow books.
• A market is not an oracle. A probability is a long-run statement, not a verdict on the next single event. A favorite priced at 80 cents still loses one time in five, and on truly novel events with murky rules the market can be confidently wrong.

How a careful person uses one

Treat the market price as a well-informed baseline, not a final answer. Form your own estimate of the probability first, from base rates and the evidence, then compare it to the price and pay attention to the gap. Before trusting any apparent mispricing, read the resolution rules closely, because many obvious edges dissolve once you see the technicality the market actually settles on. Check the market's depth and volume so you know whether the price reflects a crowd or a single trade. Size any position small, keep a record of how your calls actually turn out, and only trade where it is legal for you. None of this is financial advice, and no amount of careful reading removes the risk of being wrong; the goal is a clearer, better-calibrated view of an uncertain future, not certainty.