Prediction Market Resolution and Oracles: Read the Rules Before You Trade

Resolution is the part of a prediction market that turns a probability into a payout, and it is the part traders overlook most often. You can be right about the world and still lose if the contract's rules define the event differently from the way you assumed. The title is a question. The resolution rules are the answer.

That is why the first thing to read in any market is not the price chart. It is the fine print: who decides, from what source, by what date, and what happens if the event is ambiguous. Polymarket's own documentation warns that the market title describes the question but the rules define how it resolves.Resolution rules

The rules decide, not the headline

A typical rule set has three parts. The resolution source says where the outcome comes from, such as an official announcement, a specific website, or a data feed. The end date says when the market becomes eligible to resolve. The edge cases say what happens if the event is postponed, canceled, disputed or defined in more than one way.

Consider a market asking whether a candidate will "win the nomination." If the candidate leads after the first ballot but withdraws before the convention, did the market resolve Yes or No? The answer depends on whether the rule says "receive the most delegates" or "be formally nominated." Those are different events, and they can have different prices.

Polymarket can also publish an "Additional context" update after trading begins. The documentation says clarifications cannot change the fundamental intent of the question and are published onchain through a bulletin board contract. They are still part of the rules, and traders who do not read them can be surprised at settlement.Resolution rules

How a resolution actually happens

Most Polymarket prediction markets use the UMA Optimistic Oracle. It is fast when the answer is obvious and slow when contested.

1. A proposer selects the winning outcome and posts a bond, typically $750.

1. A challenge period, normally two hours, begins. Anyone can dispute the proposal by posting a matching bond.

1. If there is no dispute, the proposal is accepted and the market resolves. The proposer gets the bond back plus a reward.

1. If there is a dispute, a second proposal round begins. If the second proposal is also disputed, the question escalates to UMA's Data Verification Mechanism, where token holders vote.

1. An undisputed market resolves in about two hours. A disputed one can take four to six days.Resolution timeline

UMA says its optimistic oracle resolves about 99.8% of requests without a dispute. The disputes that do occur go to the DVM, where stakers vote in a 24-hour commit period and a 24-hour reveal period, with a 65% majority required for a result. Stakers who vote against the majority are slashed, and unstaking takes a week, which is intended to make corrupting the oracle costly.How UMA works

UMA's prediction-market integration adds two practical details. Proposals come from a whitelist of addresses, and the challenge period can be extended for complex or subjective questions beyond the default two hours. A second disputed request goes to a vote rather than looping indefinitely.UMA prediction markets

Not every market uses the same mechanism. Polymarket's up-or-down crypto markets use Chainlink TWAP prices rather than UMA, because the resolution is a deterministic comparison between a start price and an end price. The rule is mechanical: up if the final price is equal to or higher than the price to beat, down if it is lower. There is no proposal or vote.Resolution rules

The bond is a real incentive, and a real cost

The bond is not decorative. If you propose an outcome and you are wrong, you can lose the entire bond. If you dispute and you are wrong, you can lose yours. The docs describe the payout this way: the winner gets their bond back plus half of the other side's bond. With a $750 bond, a correct disputer risks $750 to earn $375 before costs, and a correct proposer's upside is symmetric. The risk-reward is deliberately skewed: a correct challenge earns $375, while a frivolous one loses the full bond.

That economics also explains why disputes are rare. A dispute is not a free opinion. It requires capital, evidence and time, and the market stays unresolved while the process runs. For a trader, the practical cost is that your capital is locked for days and the final outcome can differ from what the last trade implied.

Where ambiguity comes from

Most resolution problems come from one of four places.

Time. "By Friday" can mean the end of the business day in New York, the end of the calendar day UTC, or the moment a clock strikes midnight in a specific time zone. A market with an end date may resolve on the date of the event or the date of the official confirmation, which can be later.

Source. "Official" can mean a government body, a court, a company press release, or a major news wire. If the source changes its story, the rule may not say which version controls. A market that resolves on a "report" is different from one that resolves on a "confirmation."

Definition. Words like "win," "nominate," "announce," "invade," "meet" and "pass" have ordinary meanings and legal meanings. A threshold can be inclusive or exclusive. A partial outcome can be a Yes in one rule and a No in another.

Edge cases. What happens if the event is postponed, the candidate drops out, the source is unavailable, or two outcomes are arguably correct? Good rules answer these questions before trading starts. Bad rules leave them to the dispute process.

A checklist before you trade

1. Identify the resolution source. If you cannot access it, or it is not named, treat the market as higher risk.

1. Write down the exact condition for Yes. Not the headline, the condition. If you cannot state it in one sentence, you do not understand the trade.

1. Check the time zone and the end date. Convert them to your local time and note whether the rule uses the event date or the confirmation date.

1. Look for edge-case language. Search the rules for "postponed," "canceled," "withdrawn," "official," "announcement" and "if."

1. Read the clarification history. An "Additional context" update can change the practical meaning of the rule without changing the title.

1. Compare your reading with the price. If the market is at 80 cents and your reading says the rule is ambiguous, either you have an edge or you have missed something. Assume the latter until you have checked.

1. Size for resolution risk. A market with a clean rule and a public source deserves a larger position than one with a subjective standard and a history of disputes.

The limits of oracles

Oracles reduce resolution risk; they do not eliminate it. A bond can be posted by someone with a strong view, a dispute can be funded by a trader with a position, and a DVM vote is a governance process, not a court. The UMA documentation itself describes the DVM as a Schelling point mechanism, which is a sophisticated way of saying that the outcome depends on what stakers collectively believe the right answer is.

That is usually fine for clean questions. It is less fine for questions that turn on interpretation, politics or incomplete information. The market can resolve 50/50 in rare cases, and the rules allow that outcome. A trader who assumes every market resolves cleanly is taking a risk that is not reflected in the quoted probability.

The CFTC's investor education page makes the same point from the regulatory side: customers are entitled to clear information about contract terms, including payout, prices, and how settlement determinations are made, who decides, and how.CFTC explainer If a market does not give you that, the missing information is itself a risk.

Bottom line

Resolution risk is contract risk. It is not diversifiable in the same way as price risk, because a single ambiguous word can flip the payout. Read the rules, check the source and the date, look for clarifications, and size the position according to how confident you are in the wording rather than how confident you are in the event.