As you probably know if you’re reading this, Polymarket allows you to trade on the outcomes of future events. Prices between $0.00 and $1.00 reflect collective forecasts. Once an event concludes, the market must be reconciled or resolved to a binary “yes” or “no.” The outcome of this determines which traders profit and which lose.
Trust in resolution is essential. Obviously, you don't want bad actors manipulating outcomes in their favor. Because Polymarket is a decentralized protocol, they don’t rely on a central admin, like Commodus here, to decree the outcome.
Instead of someone like Commodus, Polymarket relies on rules, bonds, and third-party arbitration to align incentives.
How Polymarket Resolution Works
For a market to be resolved, an outcome must be “proposed.” To do this, the proposer submits an outcome and puts up a bond of (usually) 750 USDC, which acts as a bounty for anyone who can prove the assertion wrong.
After a market is proposed for resolution, it goes into a challenge period for two hours. If the proposal is valid and verified the proposer will receive a reward. If someone does not agree with a proposed resolution (say you don’t think Maximus should be sentenced to death) that proposal can be disputed.
If someone disagrees with a proposed outcome, they can place a dispute (along with a 750 USDC challenge bond) through the UMA dapp. UMA, or Universal Market Access, describes itself as “a decentralized truth machine,” an Optimistic Oracle (OO) that records verifiable truth onto a blockchain. The oracle is considered optimistic because it assumes honesty until challenged, and the answer stands if nobody contends the proposal.
When a dispute is placed, the case escalates to UMA’s Data Verification Mechanism, or DVM. You can think of UMA’s DVM as effectively a decentralized court.
This begins a 24-48 hour debate or evidence period. During this time, participants present sources and dispute evidence in the UMA Discord server. After the debate period, UMA token holders vote. And staked $UMA voters commit their votes in secret; this is intended to reduce collusion by preventing voters from simply copying the majority as the votes come in.
The intention is to guide voters to disregard their personal opinions and to encourage them to predict what other rational, rule-observing voters will conclude. Voters are rewarded if they vote with the (eventual) majority, and those who are inactive or vote with the minority are "slashed,” losing a portion of their staked $UMA tokens.
Once the results are in and the voting period is completed, there are four possible outcomes:
The proposer wins and receives their bond back plus half the disputer’s bond as a bounty. The disputer loses their bond.
The disputer wins and receives their bond back plus half the proposer’s bond as a bounty. The proposer loses their bond.
It is deemed too early. This outcome is for when the underlying event has not happened yet. In this case, the disputer receives their bond back plus half the prospers bond as bounty, and proposer loses their bond.
The outcome is unknown (50/50). This is rare, and is for events where none of the other options are appropriate. In this case the market price resolves to 50% yes and 50% no. The disputer receives their bond back plus half the proposer’s bond as a bounty, and the proposer loses their bond.
When the System Breaks
The trust in this process holds as long as the Cost of Corruption (CoC) exceeds the Profit from Corruption (PfC). When participants stand to lose money, they’re motivated to find fault. If they believe the rules were misapplied (and there is evidence to support them), they can dispute.
This system appears to work most of the time.
Although, in theory, someone could buy enough $UMA rig a vote, doing so would destroy the token’s value, making the attack self-defeating. The system, however, is not perfect. Issues can arise when a single market’s payoff is greater than the honest voting base. There are a few notable cases where this happened:
March 2025 — “Ukraine mineral deal” market. A high‑stakes market on whether Ukraine would agree to a Trump‑brokered mineral deal before April was resolved “Yes” despite no verified agreement. Reporting traced the outcome to a concentrated $UMA holder (“whale”) whose voting power with “as much as five million governance tokens” swung a premature DVM result. Honest voters who chose the factual outcome were slashed, which is an inversion of the intended incentives. Polymarket said they would take steps and build systems to make sure this didn’t happen again.
July 2025 — “Zelensky suit” controversy. A market on whether Zelensky wore a suit before July turned into a lightning‑rod dispute over definitions and voter concentration. “Suitgate” exposed the consequences of when subjective criteria smashes into whale‑dominated voting. You can read UMA’s debrief of the situation here.
Who gets to vote?
You do! If you want to. Any $UMA tokenholder can participate in the Data Verification Mechanism (DVM) and act as a resolver for Polymarket disputes.
The process involves three steps:
Stake: You must stake your $UMA tokens in the UMA voting portal. This locks your tokens and makes you an active participant in the DVM.
Vote: When a dispute is active, you must participate in the 48-hour commit-reveal voting process.
Claim Rewards: For staking and voting correctly, you earn a share of the protocol's inflationary rewards, as well as a pro-rata share of the penalties "slashed" from incorrect or inactive voters.
Until recently, Polymarket only relied on UMA for resolutions. However, that is starting to change.
In September 2025, Polymarket quietly began to migrate a portion of its resolution stack to Chainlink’s Data Streams and Automation, particularly for price-based markets such as the 15 minute “BTC Up or Down” and similar markets. Polymarket continues to use UMA for subjective or rules‑heavy questions where human interpretation matters.
With the rumored $POLY token set to introduce in-house oracle governance, Polymarket appears to be moving toward its own “truth layer.” While this would bring market resolution under native, token-weighted governance, it also raises new questions around decentralization, incentives, and regulatory risk.
Did you know you can still take positions on disputed contracts?
An interesting bit about UMA is that you can still buy contracts while a market is in dispute, and the public conversations on the UMA discord can provide additional insight into how the dispute might go. (Please note, you do not have to be voting to be in the conversation. And based on our own research the conversations do not seem to be very lively or involved.)
How to read a dispute before it resolves
Here’s an approach you can use to try to gain insight into which way a dispute might swing:
Start with the rules. Read the ancillary data line by line. Definitions, time windows, and exclusions usually decide the case. Disregard what traders "assumed" if it isn't written in the rules.
Write the test. If possible, in one sentence, state the criterion as a boolean: Resolve Yes iff X occurs by T as reported by S.
List admissible sources. Use only what the rules authorize (official sites, specific APIs, named outlets). Do some sources have priority over others?
Track the arguments. Follow proposer/disputer summaries and community analysis in public channels, such as the UMA Discord or X. This does not mean any of this is necessarily an admissible source, but it might give you insight into how people are thinking.
Watch the bond & timing. Rising bond sizes, repeat challenges, or “too‑early” flags might signal uncertainty.
Anticipate edge cases. Words like exclusively, consensus, recognition, suit often carry the dispute.
A mental model you might want to try: Ask, What would a rules‑maximalist majority choose using only the allowed sources?
By applying these guidelines, you can approach disputed markets more deliberately, using a clearer grasp of Polymarket’s resolution mechanics to look for, rather than assume, potential edge.
