What Is Polymarket? Prediction Markets, the CLOB and Polygon Settlement

Polymarket is a peer-to-peer prediction market. You buy and sell shares in a yes-or-no question — whether a candidate wins, a central bank cuts, or a team lifts a trophy — and each share pays $1 if the answer is yes and $0 if it is no. There is no bookmaker on the other side of your trade. Prices come from a central limit order book (CLOB) where users post bids and asks against each other, and matched trades settle on the Polygon blockchain.

A YES price of $0.63 is usually read as a 63% probability, but that number is a midpoint, not an execution price. If the best bid is $0.62 and the best ask is $0.64, a buyer pays $0.64 and a seller receives $0.62. The two-cent gap is the spread, and it is the first cost every trader meets — often a larger cost than the fee.

The CLOB: offchain matching, onchain settlement

Polymarket's order book is a hybrid. Orders are signed messages; an operator matches compatible orders offchain, then submits the matched trade to Polygon, where the exchange contract verifies both signatures and settles atomically. You keep custody of your assets until a trade settles. All orders are technically limit orders, and a "market order" is simply a limit order priced to cross the book immediately. A buy of 100 YES shares at $0.60 costs $60 and delivers 100 outcome tokens.

Every binary market has two ERC-1155 outcome tokens, YES and NO. One complete set is backed by $1 of pUSD, Polymarket's USDC-backed collateral. That backing is what makes the payoff legible: one winning token redeems for $1, one losing token for $0.

The split and merge functions are the plumbing behind that promise. Splitting $10,000 of pUSD creates 10,000 YES and 10,000 NO tokens. Suppose you sell the 10,000 NO tokens at $0.38. You receive $3,800 and keep 10,000 YES. Your net outlay is $6,200. If YES resolves true, the position pays $10,000, a profit of $3,800 on $6,200, or 61.3%. If NO wins, the YES tokens are worthless and you lose $6,200. The break-even probability is 62%, not 38%, because you financed the position by selling the other leg rather than paying for it outright.

That example also shows why the midpoint can mislead. The trade at $0.38 for NO is economically similar to buying YES at $0.62, but the executable prices include the spread and any price impact. A market that displays 60% may only let you buy at 62% or sell at 58%.

Who can use Polymarket

Access depends on location, not just on having a wallet. Polymarket's geoblock documentation lists OFAC-sanctioned jurisdictions — Iran, Syria, Cuba, North Korea, and the Crimea, Donetsk and Luhansk regions of Ukraine — as completely blocked: no new orders and no closing trades. A longer list is close-only on the global frontend and API, meaning existing positions can be closed but new ones cannot be opened. That list includes the United States, United Kingdom, France, Germany, Italy, Australia, Brazil, Poland, Singapore, Thailand, and several Canadian provinces. Ireland, Japan, the Netherlands, South Korea and Malta (sports only) are close-only on the frontend but not restricted at the API level.

US residents are not left without a venue. A separate Polymarket US venue operates a CFTC-regulated designated contract market, and the CFTC approved Polymarket's re-entry into the United States in September 2025. That is a separate exchange with its own rulebook and product set, not the global order book described in this article. The practical lesson is to call the geoblock endpoint or read the current restricted list before building a strategy around an account, because the list changes with regulatory conditions.

What resolution actually means

Every market has written resolution rules: the source, the deadline, and how edge cases are handled. Prediction markets on Polymarket use UMA's Optimistic Oracle. A proposer posts a bond, typically $750, and names the winning outcome. A two-hour challenge window follows. If nobody disputes the proposal, the market resolves in about two hours. A dispute opens a second proposal round; a second dispute escalates to UMA's Data Verification Mechanism, where token holders vote after a 24-48 hour debate period. A disputed market can take four to six days. If the outcome is genuinely indeterminate, a market can resolve 50/50, and each side redeems for $0.50.

The market title is not the contract. The rules are. That matters most for questions with subjective criteria — whether someone was "indicted," whether a policy counts as a "recession," or whether a candidate "won" before certification. Polymarket can publish additional context for unusual cases, but that context cannot change the fundamental intent of the question. Read the rules before sizing a position, not after a dispute starts.

Polymarket also lists up-down crypto markets that resolve from Chainlink time-weighted average prices rather than UMA. Those have fixed start and end observations, so the resolution path is mechanical rather than proposal-based.

The core risks

• Resolution and oracle risk. Ambiguous rules, a contested proposal, or a UMA vote can leave you with a different payout than the headline implied. A 50/50 split is possible.
• Smart contract and collateral risk. Trading relies on audited contracts, but audits are not a guarantee. pUSD is backed by USDC, so its stability depends on the issuer and the onchain backing mechanism.
• Liquidity and spread risk. Polymarket's book has no fixed trading size limit, but large orders move the price. Thin markets can show a midpoint that no one will actually trade at.
• Regulatory and access risk. Geoblocking, close-only designations, and enforcement actions can change whether you can open or close a position. A market being live today is not a promise it will be accessible tomorrow.
• Operational risk. Lost keys, phishing, and deposits on the wrong chain can be irreversible. Smart-wallet users have recovery paths, but they are not universal.

None of these risks is exotic. They are the normal risks of an exchange that settles on a public blockchain and resolves through a decentralized oracle. The difference is that Polymarket puts them in the same place: the price is a forecast with a spread, the token is a claim on $1, and the rules are the contract.

What Polymarket is not

It is not a sportsbook with fixed odds, and it is not a brokerage account with deposit insurance. It is not a polling aggregator either, although its prices are often quoted like polls. The price is a tradable opinion backed by collateral. That distinction explains both the appeal and the failure modes: you can exit before resolution, you can be right about the event and still lose money on the entry price, and you can hold a winning token that takes days to resolve.

For a trader, the useful mental model is an exchange first and a forecasting tool second. Check the book, check the rules, check the resolution source, and size the position to the liquidity rather than to the headline probability.