Prediction Markets Explained (How They Work, Legality and Accuracy in 2026)

Prediction markets went from a niche to a line on the evening news in about a year. Combined monthly trading on Kalshi and Polymarket, the two largest platforms, rose from under $5 billion in September 2025 to about $24 billion in April 2026, according to a Pew Research Center analysis of data from The Block, and then doubled again between May and July, to $53 billion a month. Newsrooms quote the odds next to polls, and the US Supreme Court has been asked to hear a case that could decide who gets to regulate them at all.

Most explainers stop at "a 62 cent price means a 62% chance" on a contract that pays $1 if the event happens. That is the right place to start and the wrong place to stop. Spreads, fees, how long your money is tied up, a taste for long shots, market depth and the written rule that decides who wins can each pull the price away from the true probability.

In short, a prediction market prices a yes or no question at between 0 and 100 cents, and the price reads roughly as a probability. In the US the CFTC regulates the exchanges while states fight their sports contracts in court, a dispute the Supreme Court may take up this winter. The research says the odds are good on liquid, near-term, clearly worded questions and much weaker on thin or distant ones.

What a prediction market is

A prediction market lets people trade contracts whose payout depends on whether a future event happens. The CFTC, the US derivatives regulator, describes the standard one as a yes or no contract with a fixed payout, usually $1, and a set expiration. Will the Fed cut rates in December? Will a given party win the Senate? If you hold YES and the answer is yes, each contract pays $1. If the answer is no, it pays nothing.

Every contract has a buyer and a seller, and the platform's business is the fee for matching them rather than a house book that wins when you lose. That is the distinction Kalshi leans on in court against states that license sports betting, and a former SEC commissioner, Joseph Grundfest, explains in a Stanford Q&A why the states call it "a distinction without a difference" when the contract is on a football game.

How the price becomes a probability

Because a winning contract pays exactly $1, its price reads as the market's probability estimate. YES at 62 cents implies roughly a 62% chance. YES and NO on the same question add up to about a dollar, since holding one of each pays exactly $1 whatever happens. Prices move when someone acts on new information. A poll lands, a court rules, a player gets injured, and the price adjusts within minutes.

The reading is approximate, though, and more so the further out and thinner the market. The price on screen is often the midpoint between the best bid (the highest price a buyer offers) and the best ask (the lowest price a seller will take), and you buy at the ask. The gap between them is the spread. Fees come on top. On most venues money tied up in a contract earns nothing until it resolves, so traders want a discount to hold it. A 95 cent contract on an event two years away pays about 5% if it wins, roughly 2.6% a year, less than cash earns in plenty of places, so it can trade at 95 cents even when the true chance is higher. That discount is why Kalshi and Polymarket have started paying interest on some positions, covered in the section on idle money below. People also tend to overpay for long shots, which the research section measures. And in a thin market one large order can move the price, so the number may reflect one trader's view more than a crowd's.

Order books and automated market makers

Kalshi and Polymarket both run order books. Traders post bids and asks, and a trade happens when a buyer and a seller agree on a price. A maker is the trader whose order rests on the book waiting, and a taker is the one who accepts it, which is what a casual buyer clicking Buy almost always does. Polymarket describes its book as hybrid, with orders matched on Polymarket's own servers and settled on the Polygon blockchain, so funds stay in your wallet until the trade executes, while Kalshi matches trades on its own regulated exchange like a futures market. Much of the liquidity on both comes from professional market makers who quote both sides and earn the spread, and Polymarket pays them daily rebates out of taker fees.

The alternative is an automated market maker, an algorithm that always quotes a price and adjusts it after every trade, such as the market scoring rule set out by economist Robin Hanson in 2002. You can always trade against it, but prices jump when the pool behind it is small, while order books give tighter prices when traders show up and empty screens when they do not.

How markets resolve and settle

Every market has written rules that define what counts as YES, which source decides it, and what happens in edge cases. The two biggest platforms enforce those rules in very different ways, and Polymarket runs two separate venues, the global polymarket.com, an offshore site closed to Americans, and a smaller CFTC-regulated exchange called Polymarket US.

On Kalshi, the exchange itself determines the outcome. Each contract names a source agency, such as a government statistics office, and Kalshi settles the market once that source publishes, which it says takes from one hour to more than twelve hours after the market closes. Polymarket US works the same way, with the exchange confirming the result from the sources named in each market's rules.

The global Polymarket hands the question to UMA, a separate crypto project whose optimistic oracle, a system that reports real-world outcomes to a blockchain, treats a proposed answer as right unless someone pays to challenge it. Anyone can propose an outcome by posting a bond, typically $750, which they lose if the answer is overturned. A two hour challenge window follows, and if nobody disputes, the market resolves. A first dispute sends the question back for a new proposal. A second sends it to a vote of holders of UMA's own token, weighted by how many tokens each holds, and the whole process takes four to six days. When a question cannot be fairly answered, a market can resolve 50/50, with every share paying 50 cents.

The money differs too. Since 28 April 2026 the global Polymarket holds positions in pUSD, its own token backed one to one by USDC, so your balance is a claim on a dollar stablecoin and inherits its risks, including the rare chance of a depeg.

Our Kalshi vs Polymarket comparison covers how the two differ for a trader, from funding to market menus, and our Polymarket explainer covers the platform itself, from fees to US access.

The 2026 landscape in numbers

Two companies dominate. Combined volume went from $26 billion in May 2026 to $53 billion in July, with Polymarket's figure including its US exchange, per the later Pew analysis. The two split volume roughly evenly as recently as March, but Kalshi has grown much faster since. Sports did most of the lifting, making up 80% of Kalshi's volume and 39% of Polymarket's in the two years from July 2024, per the earlier Pew piece.

Politics is smaller but more visible. Ahead of the US midterms on 3 November, CNN reported on 23 October that midterm markets had passed $1.4 billion in trading, nearly $1.1 billion of it on Kalshi and more than $340 million on Polymarket, per the companies. CNN, which has a data partnership with Kalshi, noted that volume counts every dollar that changes hands, so it overstates what is actually at stake.

Behind the two leaders, brokerages, sportsbooks and crypto exchanges have launched or bought their way into event contracts, while play-money sites like Manifold let people practise for free. Our platform roundup sorts the field by what each venue is actually for.

Are prediction markets legal?

At the federal level in the US, event contracts are listed on exchanges registered with the CFTC as designated contract markets, the federal licence for a US futures exchange, and the CFTC treats them as commodity derivatives. Whether that federal status stops states from policing them, mainly on sports, is what the courts are now fighting over. The dates below are as of 25 October 2026.

Kalshi has held its designation since November 2020. Polymarket's operator was fined $1.4 million for unregistered event contracts and came back through QCX, which an amended order allowed to serve US customers through brokerage firms. The global polymarket.com remains offshore, and its own list of blocked countries, updated in August 2026, still includes the United States, the UK, France, Germany and Australia among about 40 entries, with Singapore and a few others set to close-only, where existing positions can be closed but no new ones opened.

The CFTC under Chairman Michael Selig has sided firmly with the exchanges. In June 2026 it proposed rules for judging when contracts involve activities the law singles out, such as terrorism, war or gaming, and on 9 October it proposed expressly defining event contracts as swaps, calling them "within the agency's exclusive jurisdiction". The label matters because federal law gives the CFTC sole authority over swaps traded on its exchanges, so if event contracts are swaps, federal law overrides (preempts) state gambling rules for them.

The states disagree, and the federal appeals courts have split. The Third Circuit, which covers New Jersey, ruled 2 to 1 in a preliminary decision that Kalshi was likely to win its argument that federal law preempts New Jersey's gambling rules. The Ninth Circuit went the other way, letting Nevada's regulators enforce state gaming law. The Sixth Circuit, in Kalshi's suits against Ohio and Tennessee regulators, held that Kalshi's sports contracts are not swaps and that federal law would not shield them from state betting laws even if they were, the opposite of the view the CFTC set out in its 9 October proposal. New Jersey has asked the Supreme Court to take its case, backed by briefs from 39 states and the District of Columbia, the NFL and gaming groups, and ESPN reported the Court was not expected to decide whether to hear it until December at the earliest. New fights keep opening, with Utah's attorney general suing Kalshi in October.

Outside the US there is no single answer, and the platform being open in your country is not the same as using it being legal there. That question is yours to check before you fund an account.

How accurate are prediction markets?

The academic record is good, with conditions attached. The longest-running evidence comes from the Iowa Electronic Markets, a university project. Comparing its prices with 964 polls across five presidential elections from 1988, Berg, Nelson and Rietz found the market was closer to the eventual result 74% of the time, and beat the polls in every election when forecasting more than 100 days out. That benchmark is individual raw polls, though. Using the same Iowa data, Erikson and Wlezien countered that once poll leads are discounted for the time left in a campaign, poll-based forecasts beat the market. Wolfers and Zitzewitz's survey of the field still found market forecasts typically fairly accurate, beating most of the forecasting methods they were set against.

Timing matters as much as the headline. Page and Clemen found that markets are well calibrated close to expiry, meaning contracts priced at 70 cents came true about 70% of the time, but biased for events far in the future, with likely outcomes underpriced and long shots overpriced. Once you count what the money could earn elsewhere while it waits, they found, the mispricing only pays for a trader who can fund the bet cheaply, the same time-value discount described above. Overpaying for a lottery-style long shot is a cousin of the cognitive biases that cost crypto traders money.

The 2024 election gave a fresh stress test. In a 2025 working paper, Clinton and Huang studied more than 2,500 political markets on the Iowa markets, Kalshi, PredictIt and Polymarket in the final five weeks of the campaign. On PredictIt, a small US political market with capped bets, 93% of markets predicted outcomes better than chance, against 78% on Kalshi and 67% on Polymarket, so on the most heavily traded venue a third of markets did no better than a coin flip. Prices for identical contracts diverged across exchanges, and arbitrage opportunities peaked in the last two weeks before Election Day, which is exactly the kind of mispricing arbitrage traders exist to close.

The risks

The rules can resolve against what everyone assumed. In March 2025 a Polymarket market on whether Ukraine would sign a minerals deal with the US, with about $7 million traded, resolved YES although no deal had been signed, amid accusations that a large UMA holder swung the vote. Polymarket declined refunds, saying the market had not failed. You are trading on the written rule and the process that enforces it, not on the headline.

Insiders trade too. In April 2026 the Justice Department charged a US Army soldier with making about $409,881 on Polymarket using classified information about the operation to capture Nicolás Maduro. The charges are allegations until proven. In February the CFTC's enforcement division issued an advisory describing two cases Kalshi had already punished, a political candidate trading on their own race and a video editor trading on content before it went public, and warned that trading on misappropriated confidential information can break its anti-fraud rule. Manipulation is on the record as well. In July the CFTC ordered former congressman George Santos to pay about $35,000 and accept a three-year trading ban over trades on a contract asking who would attend the State of the Union.

Thin markets punish size. Polymarket's own docs warn that large orders can move the price sharply, and a market with a few thousand dollars in orders near the current price is a quote, not a consensus, as our explainer on market liquidity shows.

Fees are real, and they are highest on contracts priced near 50 cents. Polymarket's taker fee is a category rate times the number of shares times the price times one minus the price. The rate runs from zero on geopolitical markets to 0.07 on crypto markets, per its fee schedule, so 100 shares of a crypto market bought at 50 cents cost $1.75 in fees, 3.5% of the $50 stake, while makers pay nothing. Kalshi's general formula has the same shape and the same 0.07 rate, per the schedule it filed with the CFTC in 2022, though its help centre notes that some markets carry different fees and some charge makers too.

Custody and tax are on you. The global Polymarket is non-custodial, so a lost key or a drained wallet is your loss, and the usual self-custody trade-offs apply. In the US there is no settled tax treatment yet. The Bipartisan Policy Center lays out three competing views, gambling income, ordinary capital gains and the 60/40 long and short term split that US futures get under Section 1256, each taxed differently.

And between traders the game is zero-sum before fees. Every dollar a winner collects on a contract comes from another trader, so the expected value of casual trading is negative once costs come out.

What idle money earns

The platforms now pay to shrink the time-value discount. Kalshi pays a variable 3.50% APY on cash and open positions for eligible users in the United States and its territories with an average balance of at least $250. Polymarket pays a 3.25% annualised holding reward, funded from its treasury "to keep long-term pricing accurate", since interest removes the reason to discount a long-dated contract. It applies only to a short list of long-dated markets such as the 2028 presidential race and control of Congress. Elsewhere, pUSD in a position and USDC sitting in a plain wallet earn nothing.

Both programmes only help money inside those platforms. A stablecoin reserve held outside any position is a separate decision, and moving money off a platform for a few days rarely pays once withdrawal fees are counted, so it is really a question about a reserve you do not expect to trade for months. EarnPark pays up to 10% APY on USDC as of October 2026, a published rate any user gets without holding EarnPark's own PARK token. The top of that range comes from two strategies, each tagged with a Risk Label that names the kind of risk it takes. USDC DeFi is Cross-Venue, earning from hedged price gaps between trading venues, and withdraws instantly, with 0.30% of the amount plus about $3.20 charged on each withdrawal. USDC Maker Core is Market-Neutral, earning from market making hedged against price moves, and has a 30 day bonding period on withdrawals during which yield keeps accruing. These are active trading strategies rather than an interest programme, with risks of their own that the risk statement sets out. EarnPark is not available in every country, and its help centre lists the restricted regions.

What to watch next

The Supreme Court could take the New Jersey case this winter and, if it does, settle the federal versus state question as early as mid 2027. The CFTC's two proposed rules will draw comments and, eventually, final text. The 3 November midterms are the largest public test yet for political markets, and insider trading of the kind the CFTC's February advisory described is the thing to watch. Whatever the courts decide, the core mechanics will not change. A prediction market is a price on a written rule, set by people with money at stake, and it is most worth trusting when the question is clear, the market is deep and the answer is close.

If you keep a stablecoin reserve that stays out of the market for months rather than days, you can start earning on it, choosing the strategy whose withdrawal terms match how soon you might need it.

FAQ

What are prediction markets?

Prediction markets are exchanges where people trade contracts that pay a fixed amount, usually $1, if a future event happens and nothing if it does not. The contract price, between 0 and 100 cents, works as the crowd's probability estimate.

What are the biggest prediction markets?

The biggest prediction markets are Kalshi and Polymarket, which together handled about $53 billion of trading in July 2026, according to Pew Research Center, with Kalshi growing faster since March. Behind them, brokerages, sportsbooks and crypto exchanges have added event contracts, and play-money sites such as Manifold let people practise for free.

What is an example of a prediction market?

A typical example is a contract asking whether the Fed will cut rates at its December meeting. If YES trades at 62 cents, the market puts the chance at roughly 62%. Buy at 62 cents and you collect $1 per contract if the Fed cuts, or nothing if it does not, before fees.

How do prediction markets pay out?

Prediction markets pay out when the market resolves under its written rules. Each winning contract redeems for $1 and losing contracts expire worthless. Kalshi and Polymarket US credit dollars to your cash balance once the exchange confirms the result, while the global polymarket.com pays in pUSD, a token backed one to one by USDC, after the UMA oracle's challenge process confirms the outcome.

How do prediction markets decide who wins?

Prediction markets decide winners through resolution rules set before trading starts, naming the event, the deadline and the source that counts. Kalshi determines outcomes itself from named sources such as government statistics. The global Polymarket uses the UMA optimistic oracle, where a proposed answer stands unless someone disputes it, and repeated disputes go to a vote of UMA token holders.

Can you actually make money on prediction markets?

Some traders do, but between traders the activity is zero-sum before fees, so one trader's profit is another's loss. Research on long-dated markets finds their mispricing only pays for traders who can fund positions cheaply, and for most casual traders the expected return is negative once fees and spreads are counted.

Are prediction markets gambling?

Whether prediction markets are gambling is exactly what US courts are fighting over. The exchanges and the CFTC treat event contracts as regulated derivatives, while many states say sports contracts are unlicensed betting. Federal appeals courts have split on the question, and the Supreme Court is being asked to settle it.

Are prediction markets legal in the US?

Prediction markets operate legally at the federal level in the US on CFTC-registered exchanges such as Kalshi and Polymarket US, since the CFTC treats event contracts as federally regulated derivatives, while the global polymarket.com still blocks US users. Several states are challenging sports contracts in court, so what you can trade can depend on where you live and on how the litigation ends.