Between 24 and 25 March 2025, the Yes share of a Polymarket question, whether Ukraine would agree to a Trump minerals deal before April, climbed from about 9 cents to 100. The Block put the bet at 7 million dollars. The market resolved Yes although Ukraine had not formally agreed to anything. Users on X and Reddit alleged that a large holder of UMA tokens had cast enough votes to push the outcome through. Anyone asking how prediction markets work should start with that dispute. A Polymarket team member told users on Discord that this was “an unprecedented situation” and ruled out refunds, because the outcome “wasn’t a market failure”.
That one episode holds the whole subject of this lesson. A prediction market turns a yes-or-no question about the future into a share that pays 1 dollar if the answer is yes and nothing otherwise, so the share price reads as the crowd’s probability. Judging one means asking how the share is created and priced, who decides the outcome, and which regulator, if any, allows the trade.
The previous lesson, on Pendle and yield tokenization, showed a market putting a price on time. Here the market prices an event. This lesson ranks no platform and does not tell you how to bet. Prediction markets can lose you your whole stake, and in several countries they are blocked. The legal facts below change quickly: each one is dated, and every status is given as of 31 October 2026.
How prediction markets work: a share that pays 1 dollar or nothing
Yes, No and one dollar of collateral
On Polymarket, the best-known crypto-native venue, one unit of collateral creates a pair: one Yes share and one No share. The documentation calls this a split. A Yes and a No can be merged back into one unit at any time. Once the market settles, the winning share redeems for 1 and the losing share for 0. The shares are ERC-1155 tokens on Polygon.
The collateral is no longer plain USDC, which many older explanations still say. Polymarket now uses pUSD, an ERC-20 token on Polygon that its documents describe as backed by USDC, with no algorithmic peg and no fractional reserve. If you need the basics of that kind of backing, see our guide to stablecoins. The trust question does not disappear: you depend on USDC and on the contracts that wrap it.
Because a Yes plus a No always equals 1 dollar, the two prices are tied. A bid of 0.60 for Yes and a bid of 0.40 for No sum to 1.00 and can be matched to mint a new pair. That is why nobody needs a bookmaker to take the other side: the two traders fund the payout between them.
Take a worked example with invented numbers, on a market in a category with a 0.04 fee rate. You buy 100 Yes shares at 0.40, which costs 40 dollars. The taker fee is 100 × 0.04 × 0.40 × 0.60, or 0.96 dollars. If the answer is yes, you receive 100 dollars and your profit is 59.04. If the answer is no, you lose 40.96. To break even over many such trades, the event must really happen about 41 percent of the time, not 40.
The order book: signed off-chain, settled on Polygon
Polymarket runs a central limit order book, not an automated market maker. Orders are signed messages matched off-chain by an operator and settled on-chain, and every order is technically a limit order. So the venue is hybrid. Your position sits in contracts you can inspect, but the matching engine is a service that someone runs and that can apply delays, for instance around live sports.
The number shown on screen is not always a price you can get. Polymarket displays the midpoint of bid and ask, or the last trade when the spread is wider than 10 cents. In the example from its documents, a bid of 0.34 and an ask of 0.40 display as 37 percent, while a buyer pays 0.40. On thin markets that gap matters more than the fee.
For events where only one of many outcomes can win, such as an election with several candidates, a mechanism called negative risk links the markets: a No share on one outcome can convert into a Yes on each of the others. That link gives traders a reason to keep the prices of a whole event close to a total of 100 percent.
Fees are charged to takers only as of 31 October 2026, using the formula shares × rate × price × (1 − price). The rate is 0.07 for crypto, 0.05 for sports, 0.04 for politics and finance, and zero for geopolitics. The fee peaks at a price of 0.50 and shrinks toward either end. Makers pay nothing and share a rebate of 15 to 25 percent of those fees in most categories. Polymarket charges no fee of its own on USDC deposits or withdrawals, though intermediaries may.
Kalshi: the same payoff inside a regulated exchange
Kalshi sells the same binary payoff in a different legal wrapper. It describes itself as a designated contract market overseen by the CFTC, the US derivatives regulator, which means an exchange authorised to list futures, swaps and options on commodities. It runs as a company with its own trading fees, not as a set of Polygon contracts. Its contract terms name “source agencies”, such as a league or the Associated Press for a sports title, whose reports decide the result, and the exchange can open its own outcome review. The token vote described next is Polymarket’s system.
Who decides the answer: the UMA optimistic oracle
A blockchain cannot see whether Ukraine signed a deal. It needs an outside input, which is the job of an oracle, a subject covered in our advanced guide to DeFi oracles. UMA’s version is called optimistic because a proposed answer is accepted unless somebody objects. UMA says about 99.8 percent of requests across all its users end without a dispute. That figure is UMA’s own and is not specific to Polymarket.
Proposal, challenge, dispute, vote
Polymarket’s resolution page sets out the path below, as read on 31 October 2026.
| Step | What happens | Numbers |
|---|---|---|
| Proposal | A proposer picks the outcome and posts a bond | Typically 750 pUSD |
| Challenge window | Anyone may dispute; with no dispute the market settles | 2 hours |
| First dispute | The disputer posts an equal bond and a new proposal round opens | Typically 750 pUSD |
| Second dispute | Debate on UMA’s Discord, then token holders vote | 24 to 48 hours of debate, about 48 hours of voting |
| Payout | The winner of the dispute gets their bond back plus half of the loser’s | Disputed cases take 4 to 6 days |
The vote happens in UMA’s Data Verification Mechanism. Stakers commit a hidden vote for 24 hours, then reveal it over another 24. A dispute is settled once at least 65 percent of staked UMA backs one outcome. Stakers who skip the vote or land in the minority are slashed, and that stake goes to the majority. Leaving takes a one-week unstaking period. UMA’s argument is that buying 65 percent of the stake would be costly and would damage the token. That is reasoning, not a guarantee.
Voters do not judge the event in the abstract. They read the market’s written rules. Polymarket may add a clarification, which its documents say cannot change the basic intent of the question, and a market that fits neither answer can settle at 50/50, with each share redeeming for 0.50. Short crypto up-or-down markets skip UMA and settle from a Chainlink price stream.
Two disputed resolutions, and what is proven
The minerals market of March 2025 is the first case. The Block reported the claims about a large holder as allegations from users, and no finding has established them. The same report recalls that in 2024 Polymarket overturned a UMA vote on a market about Barron Trump and the DJT memecoin, which shows that the last word had not always been the oracle’s.
The second case is a piece of clothing. A market on whether Volodymyr Zelensky would be photographed or filmed in a suit between 22 March and 30 June 2025 drew more than 237 million dollars of volume, according to Decrypt. He appeared at a NATO event on 24 June in a black jacket and matching trousers, and many traders called that a suit. A first Yes was challenged, and on 1 July 2025 the process ended on No, on the ground that credible reporting had not reached a consensus. Critics blamed token-weighted voting. UMA said there was “no evidence” of foul play, according to The Defiant, and we found no published proof of manipulation.
The response came in August 2025. UMA’s blog of 12 August 2025 describes a proposer whitelist set per request. Polymarket’s default list at launch held 37 addresses with 20 or more proposals in three months and accuracy above 95 percent. Those proposers had made 96 percent of proposals, with a cumulative accuracy of 99.7 percent, against 85.8 percent for the others. The two sides do not describe the same thing, though. Polymarket’s resolution page, read on 31 October 2026, still says “Anyone can propose” and does not mention a whitelist, and we could not establish which markets use one. What UMA does state is the limit of the change: disputes stay open to all. The final vote stays weighted by tokens.
Resolution risk comes from the wording of the rules, the governance of the oracle and the incentives of the people proposing. If a market settles against you, no insurer and no house covers the loss.
Gambling, swaps or binary options: why regulators disagree
The same contract can be three legal objects. A derivatives regulator sees a swap. A state gaming board sees a sports bet. A European securities regulator sees, in some cases, a banned binary option. Which label applies decides who may offer it and to whom.
United States: a split between federal appeals courts
Polymarket’s American history starts with a penalty. On 3 January 2022, the CFTC ordered Blockratize, trading as Polymarket, to pay 1.4 million dollars and wind down non-compliant markets, for running event-based binary options without the required registration since about June 2020. In November 2024 the FBI seized chief executive Shayne Coplan’s phone and electronics, according to the New York Post. The DOJ and CFTC closed their inquiries in July 2025 without charges, as Bloomberg reported. Polymarket then announced in July 2025 that it was buying the derivatives exchange QCEX, for 112 million dollars according to Bloomberg, and on 24 November 2025 the CFTC amended the designation so US customers could reach it through intermediaries. The original site still lists the US as close-only.
Kalshi’s fight has moved through two stages. It first beat the CFTC itself over election contracts in 2024, then ended up alongside the CFTC against the states over sports. Every 2026 appeals ruling below came at an early stage, on whether Kalshi was likely to win. None is a final judgment on the merits.
| Date | Body | Decision or step | Status at 31 October 2026 |
|---|---|---|---|
| September 2024 | D.C. federal district court | The CFTC exceeded its authority in barring Kalshi’s congressional control contracts | CFTC dropped its appeal in May 2025 |
| 6 April 2026 | Third Circuit (New Jersey) | 2-1 for Kalshi: likely swaps under federal jurisdiction | New Jersey petition to the Supreme Court pending, No. 26-299 |
| 10 June 2026 | CFTC | Proposed rule on event contracts under Regulation 40.11 | No final rule found |
| 28 August 2026 | Ninth Circuit (Nevada) | 3-0 for Nevada: sports contracts likely not swaps; election contracts sent back | Kalshi rehearing request pending, no decision found |
| Late September 2026 (25 September reported) | Sixth Circuit (Ohio, Tennessee) | For the states: Ohio injunction refused, Tennessee injunction overturned | Kalshi asked the full court for rehearing on 13 October, no decision found |
The Supreme Court docket shows New Jersey filed on 2 September 2026. The NFL, the American Gaming Association, 145 tribes, Ohio with other states and former CFTC chair Gary Gensler filed friend-of-the-court briefs by 8 October. Kalshi’s deadline to respond has been extended twice, most recently on 30 October, to 7 December 2026. Semafor reported on 16 October that the justices have until January to fill this term, and that Kalshi cites the coming CFTC rules in asking the Sixth and Ninth Circuits to look again (its Sixth Circuit petition is dated 13 October). Nevada objects that those rules would not change the outcome. The Court has not said whether it will take the case.
The CFTC proposal adds a twist. According to the Congressional Research Service, it would replace categorical bans with case-by-case review, count most sports contracts as gaming and remain broadly open to them, while treating contracts on injuries, officiating calls or single plays as likely against the public interest.
European Union: binary options, gambling or crypto-assets
ESMA, the EU markets authority, answered part of the question in a public statement of 3 July 2026. Not every event contract is a financial instrument. Only those whose question relates to an underlying listed in Section C(4) to (10) of Annex I of MiFID II qualify. Those count as derivatives caught by national measures on binary options, so marketing or selling them to retail clients is banned, whatever commercial name a firm gives them, and offering them even to professionals needs a MiFID authorisation. A footnote adds that a contract may also be a bet under national gambling law or, in token form and outside MiFID, a crypto-asset under the MiCA regulation. ESMA does not sort sports or election questions into either box, and the statement leaves that assessment to firms. The enforcement against Polymarket described below has come from gambling regulators.
United Kingdom: a betting exchange by another name
The Gambling Commission wrote on 4 February 2026 that the core of a prediction market resembles a betting exchange, and that current products would appear to need a betting intermediary licence if offered in Great Britain. It told unlicensed operators not to target British consumers, and warned that operating without the right licence can be a criminal offence. Polymarket lists the United Kingdom as close-only. We found no enforcement action aimed at Polymarket there.
Is Polymarket legal in France, Italy and the rest of Europe?
Short answer for France, as of 31 October 2026: no authorisation exists, the gambling regulator treats the site as illegal, and internet providers have been ordered to block it. Polymarket lists France, Italy, Germany and Belgium as close-only, meaning existing positions can be closed and no new ones opened.
France: from geoblocking to a full block
The Autorité nationale des jeux (ANJ) turned its attention to the operator, Adventure One QSS Inc., in November 2024. After a first formal notice, the company geoblocked transactions from France. On 24 February 2026 the ANJ called prediction platforms illegal gambling sites, naming Polymarket and Kalshi and citing addiction risk, no age checks and a financial incentive to sabotage events.
On 16 July 2026 the ANJ’s president ordered internet providers to block polymarket.com. The notice says the geoblock was circumvented in practice, that no identity checks (KYC) were found for French and European users, and that the home page’s live odds promote the offer. It counts, citing Similarweb, 578,751 visits and 205,057 unique visitors in June 2026. The notice says anyone advertising such an offer can be fined 100,000 euros, and that the same penalty applies to spreading odds to promote it. That is why this lesson shows no live odds.
Polymarket said on 21 August 2026 that it had filed a petition in France against the block, without naming the court. It says it blocked trading from France in November 2024 and that the next step it saw from the ANJ was a cease-and-desist order in June 2026. That is the company’s account. Not decided: we found no ruling on that petition, and no French court decision that says whether a prediction market is legally gambling. The ANJ’s notice targets the operator, providers and promoters. We found no decision punishing a user, but that is no authorisation, and the regulator calls the offer illegal.
Italy: an ADM block and a pending appeal
Italy’s customs and monopolies agency, ADM, issued two measures on 9 and 10 July 2026: an order telling providers to redirect polymarket.com, and an entry on its list of gambling sites to block. The site has been unreachable since 27 July. ADM treats it as an unauthorised betting operator. The TAR Lazio, the regional administrative court, refused urgent relief in a decree published on 5 August and did not rule on lawfulness. Polymarket withdrew its interim request on 19 August, according to smart contract holding reserves of two or more tokens, supplied by users, that a decentralized exchange uses to execute trades automatically at a price set by a formula.">pool diverge, compared with simply holding the same tokens in a wallet. It becomes permanent when the provider withdraws.">Il Sole 24 Ore, and its appeal on the merits remains. We found no merits ruling. An earlier ban in late 2025 was lifted after a challenge, and football club Lazio has ended its Polymarket sponsorship by mutual agreement. We found no statement from Consob, the Italian market regulator.
| Jurisdiction | Authority and position | Key date | Not decided at 31 October 2026 |
|---|---|---|---|
| France | ANJ: illegal gambling, website blocked | 16 July 2026 | Polymarket’s petition; no ruling found |
| Italy | ADM: unauthorised betting, website blocked | 10 July 2026, unreachable since 27 July | TAR Lazio merits appeal |
| United Kingdom | Gambling Commission: betting intermediary licence likely | 4 February 2026 | No enforcement against Polymarket found |
| European Union | ESMA: some contracts are banned binary options for retail | 3 July 2026 | Each contract and each national gambling law |
| United States | CFTC against states, circuit split on sports | 28 August 2026, 7 December 2026 | Supreme Court review; final CFTC rule |
The ANJ also names 12 countries that have restricted or blocked prediction markets: Germany, Belgium, Romania, Switzerland, Poland, the Netherlands, Greece, Italy, Portugal, Spain, Ukraine and the Czech Republic. That is the ANJ’s list at 17 July 2026, and each country’s measure differs.
Taxes: mostly unanswered
In the United States, law firm Katten noted in February 2026 that the IRS had given no definitive guidance, and we found none as of 31 October 2026. Gains could be gambling income, with losses deductible only up to 90 percent under the 2025 tax law, or financial income, possibly under the 60/40 rule of Section 1256. Our US crypto tax guide covers the general rules. In France and Italy we found no official position from the tax authorities either. Our French crypto tax guide explains the digital-asset regime, which may or may not apply to these winnings. Keep records of every trade, and ask a tax adviser.
What the markets are used for, and what the volume says
Forecasting: useful evidence, thin sample
The strongest argument for prediction markets is that a price aggregates information from people with money at stake. A 2025 arXiv preprint found that Polymarket beat polling on the 2024 US presidential election, especially in swing states. Its authors ask for more research. It covers one election, it is a preprint, and a market that forecasts well on a heavily traded race says little about a thin market on a niche question.
Hedging: the stated purpose
Derivatives exist to transfer risk, and the CFTC’s proposal, as summarised by the Congressional Research Service, weighs hedging and pricing value in its review. In theory, a firm exposed to a rate decision or a storm could offset that exposure with an event contract. We did not find well-documented cases of companies doing so at scale, so treat hedging as a possible use, not a measured one.
Sports: where most of the money seems to be
The numbers point elsewhere. The Block’s data for August 2026, the latest full comparison we could confirm:
| Venue | July 2026 | August 2026 | Change |
|---|---|---|---|
| Kalshi | $40.1 billion | $37.17 billion | -7.3% |
| Polymarket and Polymarket US | $12.89 billion | $8.16 billion | -36.7% |
| Combined | about $52.99 billion | $45.33 billion | -14.5% |
Kalshi had about 82 percent of that total. The fall followed the World Cup of 11 June to 19 July, which supports the view that sport drives volume. An American Gaming Association spokesman told Semafor that sports make up roughly 85 percent of Kalshi’s volume. The AGA is an opponent in this fight, Kalshi’s lawyer disputed the description, and we could not check the figure. If it is close, it weakens the claim that these venues are mainly forecasting tools. Volume also needs care, as the next section shows.
Three controversies: fake volume, insiders and tampered facts
Wash trading: a Columbia estimate
A Columbia study dated 6 November 2025, by Allen Sirolly, Hongyao Ma, Yash Kanoria and Rajiv Sethi, looked for clusters of wallets trading mostly with each other. It classifies nearly 25 percent of Polymarket’s historical volume as consistent with wash trading, with a peak near 60 percent of weekly volume in December 2024 and about 20 percent again in early October 2025. By category the share is 45 percent in sports, 17 percent in elections, 12 percent in politics and 3 percent in crypto. The authors call these estimates, since no ground truth exists. The paper is a working paper and proves no intent by any named person.
Insider trading: an indictment, not a verdict
In January 2026, a recently created Polymarket account made more than 400,000 dollars betting that Nicolás Maduro would lose power, shortly before his capture, and 12 senators wrote to CFTC chair Michael Selig. On 23 April 2026, US prosecutors unsealed an indictment against a US Army Special Forces soldier, alleging that he used classified knowledge of the raid to place about 32,000 dollars of bets through pseudonymous accounts. The CFTC filed its own action, and NPR called it the first time US officials had brought criminal charges over prediction market wagers. These are allegations, not findings. He has pleaded not guilty and is presumed innocent unless a court finds otherwise. A tentative trial date of 7 December was set in June. His lawyers have asked the judge to dismiss the indictment, arguing that the charges are novel and the trades not criminal, and prosecutors opposed that in August. We found no ruling on the motion.
The tension exists whatever the verdict. A market built to gather private information pays the people who hold it, and that can be useful knowledge or a secret the holder had a duty to keep.
Weather sensors: when someone may change the fact itself
Oracle risk has a physical form. The sensor at Paris-Charles de Gaulle jumped to 22°C on 6 and 15 April 2026, moves that reportedly set the daily high on which Polymarket temperature markets settle. Météo-France complained of tampering with an automated data system. The ANJ says Paris prosecutors’ cybercrime section opened an inquiry on 4 May 2026. The cause is unconfirmed, the hair dryer story is speculation, nobody has been named, and reported winnings vary by outlet. If tampering is proven, it would fit the sabotage incentive the ANJ described in February.
Six questions to ask before you trust a price
This list is for reading a market as a source of information, for instance in a news story. It is not a method for betting.
- Read the full resolution rules, including the named source and the deadline.
- Ask whether a single person or small group could change the outcome or the data source.
- Compare bid and ask, because the displayed number may be a midpoint or an old trade.
- Check how deep the book is, since a small order can move a thin market.
- Find out who settles the market and how a dispute would work there.
- Check the legal status where you live, and treat a block as a block.
The dates to watch
Two dates fall on the same day, 7 December 2026: Kalshi’s deadline to respond at the Supreme Court and the tentative start of the soldier’s trial. The justices’ choice of cases for this term should follow by January, according to Semafor. We had found no final CFTC text on Regulation 40.11 by 31 October. In Europe, the open files are Polymarket’s petition in France and its merits appeal at the TAR Lazio.
