Polymarket Resolution 2026: Who Decides Outcomes

Updated June 2026 · Independent guide · Contains affiliate links

The moment that decides whether you get paid is not the event itself, it is Polymarket resolution, and almost nothing written about it says who actually makes the call. An undisputed market settles roughly two hours after somebody posts a bond, typically $750, proposing the outcome. A twice disputed one goes to a vote of UMA token holders and takes four to six days. This guide walks the whole lifecycle, prints what happens to both bonds in every way a dispute can end, explains the commit and reveal vote that every competing page on this query skips, and puts a measured number on how often any of it happens: 1,604 disputes across 159,447 oracle proposals. Every figure below was read from Polymarket's or UMA's own documentation, or from the paper itself, on 5 November 2026.

Polymarket Resolution 2026: Who Decides Outcomes

Key takeaways

  1. Nothing resolves the instant an event ends. Someone has to propose the outcome first, and the clock that matters starts then: a two hour challenge window, after which an unchallenged market resolves and winning shares redeem for $1 while losing shares go to zero.
  2. Proposing is not free. The proposer posts a bond, typically $750, and forfeits all of it if the proposal is wrong or submitted before the event has actually concluded. That forfeit risk is the only thing standing between you and a stranger proposing whatever suits their position.
  3. One dispute does not trigger a vote. It restarts the proposal round. Only a second dispute escalates to UMA's Data Verification Mechanism, where token holders vote, and the whole path takes four to six days.
  4. Disputes are genuinely rare. A September 2026 study of the oracle's own request history counted 1,604 disputes among 159,447 proposals, a little over 1%. The tail is the problem, not the base rate.
  5. The voters are stakers with money at risk on both sides of their own judgment. UMA resolves a dispute when at least 65% of staked UMA backs one answer, and it slashes the stake of anyone who voted with the minority or did not vote at all.

On this page

  1. How Polymarket resolution actually works, end to end
  2. The timings: when you actually get paid
  3. What a dispute does, and what it does not do
  4. Who votes when it escalates: inside UMA's DVM
  5. How often markets are actually disputed
  6. When the oracle was accused of getting it wrong
  7. Two Polymarkets, two rulebooks
  8. Washington is now asking who decides
  9. How to read a market's resolution rules before you trade
  10. Frequently asked questions
  11. Sources and when we last checked

This is a plain list rather than jump links, because guide pages on this site render section text as paragraphs. If you only read two sections, read the one on who votes and the one on reading resolution rules. Together they are the only part of this page that changes what you would do before placing a trade.

How Polymarket resolution actually works, end to end

  1. Event concludes. Nothing happens automatically.
  2. A proposer posts a bond, typically $750, and submits the outcome to the UMA Optimistic Oracle.
  3. A two hour challenge window opens. Anyone can dispute.
  4. No dispute: the market resolves. Winning shares redeem for $1, losing shares for nothing.
  5. First dispute: the disputer posts a matching bond and a new proposal round begins.
  6. Second dispute: the question escalates to UMA's DVM for a token holder vote.

Polymarket does not decide outcomes, and that is the single most misunderstood thing about it. Resolution on the international platform is handed to the UMA Optimistic Oracle, a set of smart contracts that assumes a proposed answer is correct unless somebody pays to say otherwise. The company writes the market's rules and can publish clarifications onchain, but the final call on a contested market belongs to UMA token holders rather than to anyone at Polymarket.

The flow starts with a proposal. When an event has concluded, anyone can propose the winning outcome by posting a bond, typically $750, to the oracle. If the proposal stands unchallenged it settles and the proposer gets the bond back plus a reward. If it is wrong, or merely early, the proposer loses the entire bond. Proposing too soon is the common and expensive mistake, because the oracle treats a premature proposal as a failed one.

From there the market either resolves quietly or enters a dispute path. A challenger posts a matching counter bond and the proposal is thrown out, which starts a fresh proposal round rather than a vote. Only if that second proposal is also disputed does the question escalate to UMA's Data Verification Mechanism, the token holder vote. Evidence for that vote is argued in UMA's public Discord channels, specifically #evidence-rationale and #voting-discussion, which means a market worth millions can turn on a thread you are allowed to read and post in.

One category sits outside all of this. Polymarket's up and down crypto markets resolve from a Chainlink time weighted average price instead, with no proposal and no dispute process at all. If you trade those, the oracle machinery below does not apply to you.

The timings: when you actually get paid

  1. Challenge period: 2 hours
  2. Undisputed resolution: about 2 hours after the proposal
  3. Debate period, if disputed: 24 to 48 hours
  4. UMA vote, if escalated: about 48 hours
  5. Disputed market, total: 4 to 6 days

The end date printed on a market is not the payout date, and treating the two as the same is how people end up convinced their winnings have been stolen. The end date is when the question stops trading. Payment waits on the proposal and the challenge window.

On the happy path the delay is short. Polymarket's own documentation puts the challenge period at two hours and says an undisputed market resolves about two hours after the proposal. The gap between the event ending and someone bothering to propose is the variable nobody controls, and on a small market with little money at stake it can be the longest part of the wait.

Once a dispute lands, the timetable stretches by days rather than hours. A debate period of 24 to 48 hours runs while evidence is submitted, then the UMA vote itself takes roughly 48 hours. Polymarket's docs give four to six days as the realistic total for a disputed market. That is worth internalising before you buy a contract at 97 cents on the expectation of a quick turnaround, because a dispute converts a two hour wait into most of a week with your capital locked.

What a dispute does, and what it does not do

  1. Proposer wins: proposal stands. Proposer recovers their bond plus half the disputer's.
  2. Disputer wins: proposal rejected, a new one is needed. Disputer recovers their bond plus half the proposer's.
  3. Proposed too early: same as above for the disputer. The proposer loses the bond for jumping.
  4. Unknown or 50/50, rare: every token redeems for $0.50 regardless of side.

A dispute is not an appeal against an outcome you dislike. It is an assertion that the proposed answer does not match the market's written rules, and the bond is what makes that assertion expensive enough to be worth taking seriously. Both sides post the same amount, and the winner recovers their own bond plus half of the loser's. The economics are deliberately brutal: being right pays about 50% on the bond, being wrong costs you all of it.

There are four ways it can end, and only two of them are the obvious ones. The proposer can win, in which case the proposal stands. The disputer can win, which throws the proposal out and requires a new one. The dispute can be upheld on the narrow ground that the proposal was simply too early, which also returns the disputer's bond plus half the proposer's. Or the oracle can conclude that neither listed outcome applies, which is the rare and genuinely unpleasant case: the market resolves 50/50 and every token, yes or no, redeems for 50 cents.

That last outcome deserves more attention than it usually gets. A 50/50 resolution is not a refund. If you bought yes at 12 cents you have quadrupled your money on a question that was never answered, and if you bought at 88 cents you have lost most of your stake without being wrong about anything. It is the oracle's way of saying the question was badly written, and the cost of a badly written question falls on whoever happened to be holding.

Who votes when it escalates: inside UMA's DVM

  1. Rounds turn over at 24:00 UTC. A dispute must be at least 4 hours old to join one.
  2. Commit period: 24 hours of secret voting.
  3. Reveal period: 24 hours.
  4. Resolution threshold: at least 65% of staked UMA on one outcome.
  5. Majority voters collect the slashed stake of minority voters and of non voters alike.
  6. Unstaking takes one week, which is UMA's stated cost of corruption.

This is the part the competing explainers wave at and move past. They tell you UMA token holders vote. They do not tell you how, which is unfortunate, because the mechanism is where the risk lives and UMA documents it openly.

Voting runs in fixed rounds that turn over at 24:00 UTC, and each round has two halves: a 24 hour commit period where votes are submitted secretly, then a 24 hour reveal period. The secrecy is deliberate. If votes were public, a staker's cheapest strategy would be to copy whoever voted first, and the whole thing would collapse into herding rather than independent judgment. A dispute also has to be at least four hours old before it can enter a round, so the calendar, not the urgency of your position, sets the pace.

A dispute resolves when at least 65% of staked UMA backs a single outcome. That threshold is a quorum and a supermajority at once, and it is why a genuinely ambiguous market can sit unresolved through a round. The incentive layer is sharper still: stakers who vote with the majority collect the slashed stake of everyone who voted with the minority or failed to vote at all. Not voting is punished exactly like voting wrongly.

Read that incentive twice, because it cuts both ways. It is a Schelling point design, and its intended effect is to make honest convergence the profitable strategy. Its unintended effect is that a staker's safest move is to predict what the majority will say rather than to work out what is true. When those two answers agree, which is the overwhelming majority of the time, the system is efficient. When they diverge, the design does not obviously favour truth. UMA's own counterweight is that unstaking takes a week, so anyone who corrupted a vote would be holding a depreciating token while trying to leave.

How often markets are actually disputed

  1. 1,604 disputes among 159,447 oracle proposals, a little over 1% (Nechepurenko, arXiv 2609.15368, 14 September 2026)
  2. About 99 in 100 markets resolve on the undisputed path in roughly two hours
  3. Disputes concentrate on loosely worded, high value questions rather than spreading evenly

Almost every page on this topic either implies disputes are routine, because disputes are the interesting part, or asserts they are vanishingly rare without a number. There is now a number. A September 2026 preprint by Maksym Nechepurenko reconstructed the oracle's request history and counted 1,604 disputes among 159,447 proposals, which works out at a little over 1%.

So the base rate is reassuring and the distribution is not. Roughly 99 markets in 100 resolve on the quiet path, in about two hours, with nobody paying attention. The remaining 1% are not randomly scattered either: they cluster on exactly the questions where the wording is loose and the money is large, which is to say the markets you are most likely to have an opinion worth trading. A 1% chance across all markets can be a much higher chance on the specific contract you were drawn to because its resolution criteria looked arguable.

The same paper adds one point worth knowing while you watch a pending market: the age of an oracle request says little about how far resolution has progressed, because the median wait from request to first proposal varies by orders of magnitude across the oracle's older and newer routes. A market that looks stuck may just be waiting for a proposer.

When the oracle was accused of getting it wrong

The case every serious trader should know is the Ukraine minerals market of March 2025. The question asked whether Ukraine would agree to a minerals deal with Trump before April. Around $7 million was staked. Over 24 to 25 March the yes price climbed from roughly 9% to 100%, and the market resolved yes at a point when no deal had been publicly agreed.

Traders alleged that a large UMA holder had voted enough stake to force the outcome. Polymarket's own response is the part worth remembering: it said it was aware of the alleged governance attack, acknowledged the market had resolved against expectations, called the episode unprecedented, said it was working with the UMA team to prevent a repeat, and declined to refund anyone on the grounds that this was not a market failure. Whatever you think of that reasoning, it is a clear statement of where the risk sits. The platform did not consider itself the backstop.

I would not over generalise from one incident, and the 1% dispute rate above argues against treating it as normal. But it establishes three structural facts: a token weighted vote can in principle be decided by concentrated holdings, there is no appeals process once a DVM vote concludes, and a resolution you believe is plainly wrong is not a situation in which anybody reimburses you. Price those into any contract whose wording leaves room for argument.

Two Polymarkets, two rulebooks

Everything above is documented for Polymarket International, and that qualifier matters more than it sounds. Polymarket now runs a separate US product with its own documentation site, and that documentation does not describe the proposal and dispute flow set out on this page. The two products are documented apart, and the UMA lifecycle belongs to the international one.

The instruction is narrow and worth following anyway: do not assume a market on the venue you are using resolves the way this guide describes, and check whose rules you are trading under before you need to know. Any page telling you Polymarket resolution means UMA, full stop, is describing one of two products. Our Polymarket versus Kalshi comparison covers the other half, since Kalshi settles from the source named in each contract's filed terms rather than by a token vote, and two exchanges can resolve the same apparent question differently.

Washington is now asking who decides

Resolution has stopped being a niche mechanical concern and become a regulatory one. On 21 April 2026 Senator Richard Blumenthal wrote to CFTC Chair Michael Selig about how prediction market disputes get settled, drawing a direct contrast between operators like Kalshi, whose own internal teams decide outcomes, and Polymarket's main platform, which hands the question to UMA's community voting mechanism. The letter asks what rules the CFTC actually has to ensure bets and disputes are settled on fair and clear terms, and to guard against manipulation by either the operator or an outside mechanism.

No outcome has followed that anyone should trade on. What it changes is the standing of the question. That Polymarket does not decide its own outcomes was for years presented purely as a strength, namely that no house can rule against you. It is now also being examined as a gap, namely that no accountable party is obliged to make you whole. Both readings are fair, and the practical lesson needs neither settled: the resolution mechanism is a term of the product, not a technicality behind it.

How to read a market's resolution rules before you trade

  1. Open the market and read its resolution criteria in full before funding anything.
  2. Identify the deciding source by name. If no named source exists, treat that as the warning it is.
  3. Write one sentence describing an outcome that meets the headline but fails the rule. If it is easy, expect argument.
  4. Check whether clarifications have been published for the market, since UMA voters are expected to consider them.
  5. For anything you intend to hold to resolution, assume the two hour path and plan for the four to six day one.

Almost every resolution complaint traces back to one moment: somebody traded a headline rather than the rules underneath it. The rules sit on the market's own page, they are short, and reading them takes less time than funding an account. Read them before you have a position, because once you are holding, your reading of an ambiguous clause quietly starts agreeing with your trade.

What you are looking for is specific. Name the source that decides, the deadline it has to be decided by, and the threshold that counts. Then try, in one sentence, to describe an outcome that satisfies the headline but not the written rule. If that sentence comes easily, the market has dispute risk priced into it whether or not the price shows it, and the Ukraine market above is what that looks like when it goes wrong.

Frequently asked questions

How does Polymarket resolution work?

On the international platform, resolution runs through the UMA Optimistic Oracle rather than through Polymarket staff. Once an event concludes, anyone can propose the outcome by posting a bond, typically $750. That proposal sits in a two hour challenge window, and if nobody disputes it the market resolves, with winning shares redeeming for $1 and losing shares for nothing. Polymarket writes the market's rules but does not cast the deciding vote on a contested market.

How long does Polymarket take to pay out?

About two hours after someone proposes the outcome, on the undisputed path, which covers roughly 99% of markets. The wait between the event ending and the proposal arriving is separate and is not fixed. If a market is disputed twice and goes to a UMA vote, Polymarket's own documentation gives four to six days as the realistic total: a 24 to 48 hour debate period plus roughly 48 hours of voting.

Who decides the outcome of a disputed Polymarket market?

UMA token holders, through the Data Verification Mechanism, but only after a second dispute. A first dispute simply restarts the proposal round. In the DVM, stakers vote in a 24 hour secret commit period followed by a 24 hour reveal, and a dispute resolves when at least 65% of staked UMA backs one outcome. Stakers who vote with the minority, or who do not vote, have stake slashed and redistributed to the majority.

What happens if a Polymarket market resolves 50/50?

Every token redeems for 50 cents, whichever side you were on. It is the oracle's verdict that neither listed outcome applies, usually because the question was written ambiguously, and it is rare. It is not a refund: a holder who bought at 12 cents profits and one who bought at 88 cents takes a large loss, with neither result reflecting whether they read the event correctly.

Can Polymarket overrule the UMA oracle?

Not as a matter of ordinary process. Polymarket writes market rules and can publish clarifications onchain that UMA voters are expected to consider, but the DVM vote is the final step and there is no appeals process once it concludes. In the March 2025 Ukraine minerals case, Polymarket acknowledged the market had resolved against expectations, said it was working with UMA to prevent a repeat, and declined to refund traders on the basis that it was not a market failure.

How often are Polymarket markets disputed?

A little over 1% of the time. A September 2026 preprint reconstructing the oracle's request history counted 1,604 disputes across 159,447 proposals. The base rate is low, but disputes are not spread evenly: they concentrate on high value markets with loosely worded resolution criteria, so the risk on a specific arguable contract can be considerably higher than 1%.

Can I dispute a Polymarket resolution myself?

Yes, during the two hour challenge window, by posting a counter bond matching the proposer's, typically $750. Understand what you are buying: a dispute argues that the proposed answer does not match the market's written rules, not that you dislike the result. If you are right you recover your bond plus half the proposer's, and if you are wrong you lose all of it, so the expected value is poor unless the rule text is clearly on your side.

Do all Polymarket markets use the UMA oracle?

No. Polymarket's up and down crypto markets settle from a Chainlink time weighted average price with no proposal and no dispute process. Separately, Polymarket's US product is documented on its own site, and that documentation does not describe the UMA proposal and dispute flow, so do not assume a market resolves this way without checking which venue and which market type you are trading.

Sources and when we last checked

  1. Polymarket documentation, the Resolution concepts page and the developer UMA resolution lifecycle page, both read 5 November 2026: the typically $750 bond, the two hour challenge period, the 24 to 48 hour debate period, roughly 48 hours of voting, the four to six day disputed total, the bond distribution outcomes and the 50/50 case.
  2. Polymarket help centre, article 13364518: the $750 bond, the two hour challenge window and the $1 payout on winning shares.
  3. UMA protocol documentation, How does UMA's oracle work: the 24 hour commit and 24 hour reveal rounds turning over at 24:00 UTC, the four hour minimum age for a dispute to join a round, the 65% staked UMA threshold, the slashing of minority and non voters, and the one week unstaking period.
  4. Maksym Nechepurenko, Resolution Is Not Settlement, Part I, arXiv 2609.15368, submitted 14 September 2026: the 1,604 disputes among 159,447 proposals and the request age finding.
  5. The Block, 26 March 2025: the Ukraine minerals market, the roughly $7 million staked, the price move over 24 to 25 March and Polymarket's statement that it would not refund traders.
  6. Letter from Senator Richard Blumenthal to CFTC Chair Michael Selig, 21 April 2026: the contrast between Kalshi's internal resolution teams and Polymarket's UMA community voting mechanism, and the questions on manipulation.

Two transparency notes. The Blumenthal letter PDF on senate.gov and The Block's report both returned HTTP 403 to an automated fetch while this was written. The Senate letter is therefore described from Blumenthal's published press material and coverage of it rather than from the PDF text; The Block's figures were read from the article itself by another route.

One source conflict is worth naming rather than hiding. At least one widely ranking guide on this topic puts the proposal bond at $500. Polymarket's own documentation and help centre both say $750, so that is the figure used here, and Polymarket's own word is typically, meaning it can change. Check the live bond rather than trusting any guide, this one included.

Nothing here is financial advice. Carry away one line: Polymarket resolution is a term of the product rather than a formality after it, so read the rules while you can still decide not to trade.

Related guides