How do prediction markets work?

Every election season, the same argument breaks out: can you trust the polls? In recent years, a new contender has entered the chat — prediction markets, where thousands of strangers bet real money on outcomes, and the resulting prices get quoted like forecasts. Sometimes they beat the pollsters. Sometimes they do not. Either way, they are worth understanding.

Short answer: a prediction market lets you buy and sell contracts tied to a yes-or-no question — "Will the Fed cut rates in December?" Each contract costs between $0 and $1 and pays $1 if the answer is yes. The price is the market's implied probability, set by traders, not by a bookmaker. It is part exchange, part casino, part forecasting tool — and the proportions depend on who you ask.

The basic trade

Take a concrete example. The question is "Will the Federal Reserve cut interest rates at its December meeting?" You believe the answer is yes, so you buy Yes contracts at 63 cents each. If the Fed cuts, each contract pays $1 — you make 37 cents per contract. If the Fed holds, the contracts pay nothing, and you lose your 63 cents.

Someone on the other side believes the answer is no. They buy No contracts at 37 cents, which pay $1 if there is no cut. Every trade has two sides, and the prices always sum to roughly $1 — the market's way of saying the probabilities must add to 100%.

You can also sell before the event resolves. If new data arrives and Yes climbs to 80 cents, you can sell your 63-cent contracts for a profit without waiting for December. That is what makes it a market rather than a bet: prices move continuously as information arrives, and you can trade in and out like a stock.

Why the price is the forecast

When Yes trades at 63 cents, traders are collectively saying there is about a 63% chance of a rate cut. That is not a metaphor — it is arithmetic. If you genuinely believed the chance was 80%, you would buy at 63 cents all day, and your buying would push the price up. If you believed it was 40%, you would sell, pushing it down. The price settles where the marginal trader thinks the odds are fair.

This is the mechanism that makes advocates call prediction markets "truth machines." Unlike a poll, which asks people what they think, a market asks people to put money behind what they think. Talk is cheap; 63 cents is not. The theory is that financial skin in the game filters out noise, punditry, and wishful thinking.

The theory has something to it. In the 2024 U.S. election, prediction market prices moved faster than polling aggregators at several key moments, and got significant attention for it. But a price is still just an aggregation of opinions — weighted by wallet size, distorted by fees, and vulnerable to the same herd behavior as any market.

The two big platforms

The market is dominated by two venues with opposite philosophies. Kalshi, founded in 2018, is a federally regulated exchange overseen by the Commodity Futures Trading Commission — the same framework that governs futures and options. It is KYC-compliant, dollar-denominated, and built for U.S. retail and institutional users who want event exposure without touching crypto.

Polymarket, launched in 2020, is crypto-native: it runs on the Polygon blockchain, settles in USDC, and uses a central limit order book where anyone can place limit or market orders. Its global platform historically operated outside U.S. regulation, but it has since established a federally regulated U.S. operation — Polymarket US — after acquiring the CFTC-licensed exchange QCEX for a reported $112 million.

The two are converging on regulation but diverging on audience. Kalshi leans into politics, economics, and climate contracts for a brokerage-like experience. Polymarket leans into volume and liquidity, particularly around politics and crypto, for a faster-moving crowd. Most serious traders keep accounts on both and place each trade wherever the price is better.

Regulated exchange or sportsbook?

Here is where it gets legally interesting. A prediction market and a sportsbook can cover the same football game, but they are different businesses: a sportsbook sets the line and takes the other side of your bet, while a prediction market just matches buyers with sellers and takes a fee. The platform has no opinion on the outcome.

Regulators have not always seen it that way. Kalshi won a landmark legal battle in 2024 that allowed it to list election contracts, which the CFTC had previously resisted. Since then, the fight has moved to the states: gaming regulators from New York to Nevada issued cease-and-desist orders through 2025 and 2026, arguing these are unlicensed sportsbooks; the platforms sued back in federal court, arguing CFTC jurisdiction preempts state gambling law. In April 2026, the CFTC itself sued Arizona, Connecticut, and Illinois to defend federal jurisdiction over event contracts.

The outcome matters for users because it determines what can be listed, where, and with what consumer protections. A federally regulated exchange has clearing, custody, and oversight rules. An offshore crypto platform has transparency and code. They are not the same safety profile, and the marketing rarely dwells on the difference.

What the volume numbers say

This is no longer a niche academic curiosity. Combined monthly trading volume on Kalshi and Polymarket rose from under $5 billion in September 2025 to about $24 billion in April 2026, according to a Pew Research Center analysis. Trailing-year figures put Kalshi around $39.7 billion and Polymarket around $36.2 billion — a narrow gap at the top.

One number reframes the whole story: roughly 87% of Kalshi's trailing-year volume was sports. Whatever the forecasting-theory debates say, the actual business — at least on Kalshi — looks a lot like a sportsbook competitor with better plumbing. Politics and economics get the headlines; sports pay the bills.

Are they truth machines?

The honest answer is: sometimes, on some questions, better than the alternatives — and the exceptions matter.

Prediction markets work best on questions with clear resolutions, short time horizons, and enough traders to keep prices honest. Elections, Fed decisions, sports outcomes. They work worst on vague questions ("Will the economy improve?"), long-dated ones (where a dollar today is worth more than a dollar at resolution), and thin markets where a single large trader can move the price.

Manipulation is a real concern, not a theoretical one. A trader with deep pockets can push a price to create a false signal — and in 2024, there were widely discussed cases of large bets moving election markets. Platforms have surveillance and position limits, but a market is not a lie detector. It is a weighted average of what people with money believe, or want others to believe.

There is also the uncomfortable question of insider trading. If a campaign staffer trades on nonpublic polling, is that savvy or illegal? The rules are still being written, and the answer may differ between a CFTC-regulated exchange and a crypto platform.

The risks of playing

If you trade on prediction markets, understand what game you are in. This is speculation with a short fuse: most contracts resolve to $0 or $1, which means most trades are binary — you are either right or you lose everything you staked. There is no partial credit.

Fees eat into small edges. Taker fees run up to a few percent depending on the venue and the price, and frequent trading compounds them. The bid-ask spread on thin markets is a hidden tax. And unlike stocks, there is no underlying asset growing over time — every dollar you make is a dollar someone else lost, minus the platform's cut.

The people who do well tend to be informed specialists: they know a niche (a sport, a regulatory process, a local election), they size positions carefully, and they treat it as trading, not entertainment. The people who do badly treat it like a casino with better branding.

How a market decides who was right

A prediction market is only as good as its resolution process — the mechanism that decides whether Yes or No won. Get this wrong and the whole thing falls apart, because traders will not risk money on a contract that might not pay out fairly.

On Kalshi, resolution is straightforward: the exchange designates an official source in the contract terms — the Associated Press for elections, a government agency for economic data, the league for sports — and settles accordingly. It is centralized, but it is clear, and there is a regulated entity to complain to.

Polymarket's global platform does it differently. Outcomes are settled through UMA, a decentralized oracle protocol where token holders vote on the result and can be challenged. It works, but it has produced genuinely disputed resolutions — markets where the wording was ambiguous and both sides claimed victory. Anyone who has watched a close UMA dispute knows that "decentralized truth" can be messier than an official source.

This is worth checking before you trade: read the resolution criteria, not just the question. The question asks what you think will happen. The criteria decide what counts as happening. When those two diverge — a candidate "wins" but the AP has not called it, a rate cut happens but outside the specified meeting — the criteria win, and your correct prediction can still lose money.

Prediction markets are a genuine innovation in how information gets priced — and a genuinely efficient way to lose money if you confuse a price with a prophecy. The market is not telling you what will happen. It is telling you what people willing to risk money think will happen, which is useful, fallible, and occasionally manipulated. Read it the way you would read a poll with a wallet attached: informative, never final.