Prediction markets: how they really work, the complete guide

In May 2026, Kalshi and Polymarket together handled $24 billion of volume in a single month, up from $5 billion a year earlier, according to a Pew Research Center analysis. Elections, central bank decisions, matches, film releases, the weather: anything with a verifiable answer can now be traded.

From a distance, these markets look like betting sites. Up close, they work like a stock exchange: an order book, prices moving continuously, positions you can sell before the end. Here is how a contract is born, how its price forms, how it settles, and what really sets them apart from a bookmaker.

Prediction markets: the 2-minute refresher

A prediction market is a platform where people trade contracts whose value depends on the outcome of a future event. The oldest still running, the Iowa Electronic Markets, was launched in 1988 at the University of Iowa. Modern platforms, Kalshi (regulated in the US by the CFTC) and Polymarket (on blockchain), have popularised the format since 2020.

The logic is that of a discounted dollar: you pay $0.62 for a security that will be worth $1 or $0. If you are right, the gain is $0.38 per contract, or +61%. If you are wrong, the loss is total.

Step 1: how a contract is created

Everything starts with a question that must meet three conditions: a binary or multiple-choice outcome, a resolution date, and a source of truth designated in advance.

The wording is the heart of the work. "Will Trump win?" is too vague. "Will the Republican candidate be declared the winner of the US presidential election by the Associated Press before 31 December 2024?" is a contract. The rules also specify what happens in case of cancellation, postponement or ambiguity.

On a centralised platform such as Kalshi, contracts are created by the platform itself, which is a designated contract market (DCM) supervised by the CFTC: each category of contract is filed with the regulator, which can object. On Polymarket, markets are proposed by the platform and sometimes by the community, and each market is deployed as a smart contract on the Polygon blockchain, collateralised in the USDC stablecoin.

Once the market is created, two securities exist: a Yes and a No. Together they are always worth exactly $1: at any time, one Yes plus one No can be exchanged for $1 of collateral, and vice versa. This is what guarantees that prices add up to $1 and that winners will be paid.

Step 2: how the price forms

This is where a prediction market parts ways with a classic bet. Nobody sets the odds. The price results from buyers meeting sellers, just like a stock.

The order book

Kalshi and Polymarket run a continuous double auction: an order book where everyone posts the price at which they want to buy or sell Yes. The best buyer and the best seller form the spread. When a buy order meets a sell order, the trade goes through and the last price becomes the displayed probability.

Example. Contract "Bitcoin trades above $150,000 before 31 December". The order book shows:

The spread is 2 cents. If you buy 500 Yes at market, you pay $0.40, or $200. If news breaks and buyers rush in, the sell side empties and the price rises to $0.45, then $0.50. Your position is then worth $225, then $250, and you can sell it without waiting for expiry.

One detail matters: selling a Yes and buying a No are the same thing. If Yes trades at $0.40, No trades at $0.60. The platform automatically matches a Yes buyer at $0.40 with a No buyer at $0.60: together they deposit $1, which will be paid to the winner.

The automated market maker

For thinly traded markets, some platforms use an algorithm that quotes continuously, the automated market maker. The best known, Robin Hanson's LMSR rule, adjusts the price with every purchase: the more Yes is bought, the more expensive Yes becomes, along a predefined curve. The operator provides liquidity and accepts a known maximum loss in exchange.

What moves the price

Three forces: new information (a poll, a press release, a leak), sentiment, and liquidity. In a deep market, a $10,000 order barely moves the price. In a thin one, it shifts it by ten points. That is why the price of a small market is a fragile signal.

Step 3: how the contract settles

On the resolution date, the market closes. The outcome is determined according to the source defined at creation. Winning contracts settle at $1, losing ones expire at $0.

On Kalshi, the platform determines the outcome and settles, under CFTC oversight, with an internal dispute procedure.

On Polymarket, the result is brought by an oracle, the UMA protocol. A participant proposes an answer by posting a bond. During a challenge window, anyone can dispute it by posting a bond too. In case of a dispute, UMA token holders vote. Once the result is final, the smart contract releases the collateral to the winners, who claim their payout.

This mechanism has its limits. In 2025, several Polymarket markets saw contested resolutions, where the letter of the rule and the spirit of the question diverged, and where the token holders' vote settled the matter. That is the price of decentralisation: no central authority, but no neutral referee either.

The same event, two ways to play it

Let us compare a bet with a bookmaker and a position on a prediction market, on a match where team A is the favourite.

A worked example on margin. The bookmaker shows A at 1.60 and B at 2.50. Implied probabilities: 62.5% and 40%, total 102.5%. The extra 2.5% is its margin, taken whatever happens. On a prediction market, Yes A trades at $0.61 and No at $0.39: total 100%, and the cost is limited to the stated fees and the spread.

The real difference is structural: the bookmaker is your opponent, the prediction market is a trading venue. The former wins when you lose. The latter wins when you trade.

Real-world case: the 2024 US presidential election

The November 2024 election was the turning point. Polymarket recorded more than $3.6 billion of volume on the presidential winner market alone, according to its public data. From October, the market had Donald Trump as a clear favourite, around 60%, while national polls called the election a coin flip.

Two readings coexist. Supporters see it as proof that money aggregates information better than opinion polls. Critics point out that a French trader, nicknamed "Théo" by the press, had placed more than $30 million on Trump across several accounts, enough to move the price on his own in a market of that size, and that a price can be right for the wrong reasons.

The lesson applies to every market: a price reflects what those who trade know and believe, weighted by their capital. When one player accounts for a significant share of the volume, the price also becomes a measure of their conviction.

Where regulation stands

In the United States, the CFTC supervises prediction markets as derivatives markets. In 2024, Kalshi won in court the right to list contracts on elections. Sports contracts remain disputed between the federal regulator and the states, which often treat them as bets falling under their gambling laws.

In Europe, there is no dedicated framework. An event contract is, depending on the case, a derivative (MiFID II), a bet (national gambling law), or a crypto-asset. Polymarket is not available to residents of France, the United Kingdom, Germany or Belgium, among others. In France, the national gambling authority (ANJ) has reminded that only licensed operators may offer bets, and that a prediction market accessible from France without a licence is illegal.

Advantages of prediction markets

• A price readable as a probability, updated continuously.
• A position you can sell before the outcome, to lock in a gain or cut a loss.
• No bookmaker on the other side: the platform does not win when you lose.
• An information tool: prices serve as a barometer for the media and researchers.
• A maximum loss known at purchase: the price paid.

Disadvantages and risks

• Total loss on the losing contract: no partial recovery.
• Uneven liquidity: on small markets, spread and slippage are expensive.
• Possible manipulation: a large participant can move the price, and an insider can trade on private information.
• Contestable resolution: the letter of the rule can prevail over the spirit of the question.
• Uncertain legal framework and access banned in many countries, including France.
• Drift towards gambling: the fast, binary mechanics encourage overtrading.

Prediction market, binary option, stock: the comparison