Article

Inside Polymarket: UMA Resolution and Market-Making Risks

Who settles Polymarket outcomes, and how do market makers earn? Understand UMA disputes, the Ethereum ETF case, liquidity rewards and the risks behind them.

By VARVAR · Published Nov 8, 2026

A magnifying glass verifies a paper ledger between two stacks of market orders.

Polymarket uses UMA’s oracle to resolve outcomes under each market’s rules. Market makers supply trading liquidity and may earn spreads and incentives, but they also take losses. Together, these systems explain why reading the news correctly—or collecting rewards—does not guarantee a profitable trade.

From the outside, Polymarket looks almost suspiciously simple. Pick an event. Buy YES or NO. Wait. Eventually, you either feel like a financial visionary or explain that the missing money was your research budget.

Under that interface sit two different systems. One determines what counts as a winning outcome. The other gives you someone to trade with before the answer is final. Understanding both is more useful than discovering their existence while staring at an unexpected loss.

Unlike a traditional fixed-odds bookmaker, Polymarket matches participants’ orders. In an ordinary binary resolution, the winning share pays $1 and the losing share pays $0. A price around $0.65 can be read as a market-implied chance near 65%, but it is neither a guarantee nor a poll of everyone’s opinion. The price available for your order may also differ from the number on the chart.

The platform’s trading collateral is pUSD, a USDC-backed token on Polygon. You do not need a tour of the plumbing to follow the important question: what has to happen for your share to pay out?

Why the headline can be right and your trade wrong

Imagine a fictional market asking whether a rocket will reach orbit before the end of the month. You buy $1,000 of YES shares.

The rocket launches. The footage is magnificent. The internet applauds. You start researching what to buy with the winnings, which is always an efficient use of money you do not yet have.

Then the market resolves NO. The rocket lifted off, but it never reached orbit. The rules required the second event; the spectacular video showed the first.

A blockchain cannot watch the launch and interpret the flight data on its own. An oracle provides a way to bring that outside information into the contract. It still needs a question precise enough to answer.

How UMA resolves a Polymarket market

UMA’s system is called optimistic because a proposed answer can stand unless someone challenges it. This is procedural optimism, not a glowing review of human nature. Proposals and challenges come with money at risk.

The process currently described by Polymarket works like this:

  1. Propose an outcome. A participant supplies an answer and posts a bond, typically 750 pUSD.
  2. Allow a challenge. The usual window is two hours. An unchallenged proposal can be accepted.
  3. Start a second round if disputed. A challenger also posts a bond. The first dispute leads to a new proposal round.
  4. Escalate a second dispute. If that proposal is challenged too, UMA’s Data Verification Mechanism, or DVM, handles the question through a vote.

Polymarket gives approximate timelines of two hours after an undisputed proposal and four to six days for a fully disputed resolution. These are guides, not a guaranteed appointment with your payout.

YES and NO are not the only possible responses. “Too Early” means the question is not ready for final resolution. In a rare Unknown/50-50 resolution, both sides can redeem for $0.50 each. That is a specific outcome, not an automatic refund whenever traders disagree.

A proposed outcome can be accepted without a dispute. The first dispute starts a new proposal; a second dispute leads to a UMA vote.
The first challenge starts another proposal round. UMA voting is the escalation route after a second disputed proposal.

What voters actually decide

The question is not simply whether an event happened in the broad, everyday sense. It is whether the evidence meets the market’s stated conditions: the exact wording, description, named sources, deadline, time zone and treatment of unusual cases.

“The rocket flew” may be true. “The rocket reached the orbit specified in these rules before this deadline” is a different claim. The more money involved, the more impressive a career a single verb can have.

Polymarket can also publish additional context after trading begins. Its documentation says these clarifications are recorded onchain and must preserve the question’s fundamental intent. That constraint matters, but it does not make every borderline interpretation obvious.

If a question reaches the DVM, participants vote using staked UMA tokens. They first commit a concealed vote, then reveal it. A normal round has a 24-hour commitment phase followed by a 24-hour reveal phase, with participation and agreement thresholds governing whether it can finish.

This is not one person, one vote. More staked tokens mean more voting weight. Buying more YES shares does not give you a larger UMA ballot, and a furious comment is not a governance token.

The important limitation follows from that design: a decentralized process does not necessarily distribute influence evenly. That alone is not proof of manipulation. It is a reason to understand how a disputed answer becomes final rather than assuming an impartial human referee will step in.

The Ethereum ETF dispute: what did “approved” mean?

A real market in May 2024 asked whether an Ethereum ETF would be approved by May 31. It named the SEC as its primary resolution source and used a deadline of 11:59:59 PM Eastern Time. The market recorded more than $13.2 million in trading volume—not $13.2 million in losses.

On May 23, the SEC approved exchange rule changes filed under Form 19b-4 for spot Ether exchange-traded products. Making the issuers’ registration statements effective was a separate step.

That distinction was substantial. For example, the iShares Ethereum Trust ETF’s quarterly filing records its S-1 becoming effective on July 22, followed by its Nasdaq listing on July 23. The exchange-rule approval and the start of trading were different milestones.

The argument can be expressed in two sentences. One reading treats the SEC’s May approval as satisfying the market. Another asks whether an ETF could really count as approved before it was ready to trade.

The proposed YES result was disputed, and the market ultimately resolved YES. The useful lesson is not that “approved” always means the same thing. It is that a market about approval may be asking something different from a market about the first day of trading.

You can follow the news accurately and still need to reread the contract. Apparently, one hobby was not enough.

The May 2024 Ethereum ETF dispute contrasted approval of 19b-4 exchange rule changes with an S-1 registration statement not yet being effective. The market resolved YES.
The regulatory milestones were separate. The historical market’s final outcome was YES.

Five checks before you trade the wording

  1. Read the full rules. The title introduces the subject. The rules specify what the contract will pay for.
  2. Check the resolution source. It may be an official release, a named exchange, a particular price index or a consensus of credible reports. A Bitcoin price on one venue cannot settle a question that names another feed.
  3. Inspect the verbs. Announce, sign, take effect, launch, begin trading and complete describe different milestones. Do not silently substitute the one you remember from a headline.
  4. Check the time zone. A deadline in Eastern Time is not the same moment as the same clock reading in UTC.
  5. Recognize a dispute trade. Prices may start reflecting expectations about how the rules will be interpreted. You may be able to reduce or close a position while trading remains open, but only if there is a buyer at an acceptable price.

Once the argument becomes the main driver of price, you are forecasting the resolution process as well as the event. That is another layer of uncertainty, not a beginner’s shortcut.

Can market makers earn without predicting the winner?

Suppose you would rather provide prices than make one big call about a film award or a game release. You can act as a market maker: place orders that other traders can buy from or sell into.

You still need a view of fair value. If YES looks worth roughly $0.50, you might offer to buy it for $0.49 and sell it for $0.51. Buy a quantity at the lower price and sell the same quantity at the higher price, and the gross difference is two cents per share.

The word gross is doing useful work. Both trades have to happen at those prices. One fill on its own leaves you with exposure instead of a completed spread.

You also need to fund the orders: money for the bid and the corresponding shares for the sell order. The ability to place two prices is not permission to sell unlimited shares you do not own.

Market makers make trading easier by supplying available orders. In return, they seek compensation for accepting inventory, timing and price risk. The job is less “earn without forecasting” and more “keep offering prices without being run over by new information.”

A market maker offers to buy YES at 49 cents and sell at 51 cents. Both fills are needed for the two-cent spread; rewards and rebates have separate eligibility rules.
A two-cent quoted spread is only potential revenue. Unmatched fills leave a position exposed to price changes.

Liquidity Rewards: payment for qualifying orders in the book

In incentivized markets, Liquidity Rewards pay for eligible limit orders resting in the order book. Qualifying size, distance from the adjusted midpoint and two-sided quoting affect the score. Each market sets its own minimum size, maximum qualifying spread and reward allocation.

The pool is shared according to relative scores. A larger or more competitive order may earn more points, but other makers’ activity affects your share. The advertised pool is not your personal daily income.

The rules checked on September 24, 2026 allow reduced scoring for one-sided liquidity when the adjusted midpoint is between $0.10 and $0.90, inclusive. Outside that range, liquidity must be two-sided to score. Rewards are distributed daily at midnight UTC, with a $1 minimum payout.

This is payment for making capital available on stated terms, not interest on an untouched deposit. Other traders are allowed to use your orders. Occasionally, they will be delighted with the bargain.

Maker Rebates: a separate payment for orders that fill

Maker Rebates are tied to eligible maker orders that actually execute. They are funded by a portion of taker fees in qualifying markets. Your allocation depends on the eligible trading activity you provided, rather than simply how much money you deposited.

The current program pays daily in pUSD and requires at least 1 pUSD in accrued rebates for a payout. Terms differ by market and can change.

Makers currently pay no maker trading fee. That does not make every limit order free: an order that immediately takes available liquidity can be a taker order. Nor does every market charge taker fees or qualify for the same incentives.

So there are three possible revenue components: spread, Liquidity Rewards and Maker Rebates. They reward different activity. Add them together and you have revenue to compare with losses and costs—not a profit guarantee.

Adverse selection: when someone likes your price too much

A market maker’s uncomfortable question is why the other trader chose to take that quote, at that moment.

Sometimes the reason is ordinary trading demand. Sometimes the other side has reacted to information before your order has caught up. That second problem is called adverse selection.

Imagine you leave an order to buy NO at $0.49 while the event looks roughly 50/50. New information arrives, and the market’s assessment of YES jumps to 95%. NO is now valued near five cents. Your old bid is still sitting there, offering forty-nine.

A faster trader sells into it. You receive NO shares at $0.49 while the market values them near $0.05: a $0.44 drop in value per share relative to what you paid.

That is a valuation loss, not a claim that you have already sold at five cents. An actual exit needs an available bid; the price could change again, and the shares could eventually pay nothing. The attractive reward counter does not remove any of those possibilities.

In a hypothetical news shock, an old 49-cent bid for NO fills after NO is valued near five cents, leaving a 44-cent loss in market value per share.
An illustrative stale-order fill. The price comparison shows a loss in market value; realized proceeds depend on an actual exit or final resolution.

This is also why “impermanent loss” is not the best label here. That term usually concerns the changing asset mix of an automated market-maker pool compared with holding the assets. In this order-book example, the immediate problems are a stale quote and the position acquired through it.

Why the rewards may not cover the damage

Consider a separate hypothetical day. Completed trades earn $20 in spread. Liquidity Rewards add $15, and Maker Rebates add $5. So far, $40.

Then you close a different position acquired through a stale quote at a $200 loss. The combined result is minus $160, before any other expenses. This example treats that loss as separate from the profitable trades, so it is counted only once.

Small, frequent payments do not automatically compensate for occasional large losses. A strategy can show a cheerful stream of rewards while its total value falls. The spreadsheet needs the positions as well as the confetti.

Four risks that incentive payments do not remove

Inventory risk. Fills can leave you holding too much YES or too much NO. As the imbalance grows, more of your result depends on the event itself. You may become a directional trader with a more complicated screen.

Stale-quote risk. A sensible price can become a bad one after a data release, a game starts or an official announcement arrives. Known news times give you a chance to withdraw or expire orders beforehand; unexpected news does not send a calendar invitation.

Resolution risk. A market maker also needs to understand what the rules will count. Providing liquidity does not exempt you from reading the contract.

Technical risk. A bot can stop, a connection can fail, and data can arrive late while orders remain available to other traders. Position limits, checks on prices, current market data, monitoring of fills and an emergency cancellation procedure all matter.

Polymarket’s market-making documentation recommends these kinds of controls, including expiring quotes before known catalysts and cancelling stale orders. Sending a cancellation request is not the same as knowing it succeeded. Automation can carry out a bad decision very efficiently.

A beginner’s approach to market making

  • Start with a small amount of capital.
  • Avoid leaving unmanaged quotes through expected news events.
  • Check the reward settings for the specific market.
  • Do not treat the entire reward pool as your future earnings.
  • Track your remaining positions and total profit or loss, including changes in the value of unsold shares.
  • Set limits on position size, imbalance and acceptable loss before quoting.
  • Know how to stop the automation and confirm which orders are still open.

Market making is not passive income. It involves maintaining prices, watching exposure and deciding when the available compensation no longer justifies the risk.

Read the rules before counting the rewards

Resolution determines which outcome the contract pays for. Liquidity determines what you can buy or sell along the way. A good forecast can still meet an unexpected interpretation, and a well-paid order can still be a bad trade.

For a trader, read the rules before pressing Buy. For a market maker, understand the exposure before projecting the rewards. Both habits are less exciting than watching the balance move. They are also cheaper to learn in advance.

This article explains the mechanics rather than recommending a trade. Access depends on your jurisdiction; market rules, fees and incentive terms should be checked before you use them.

Your privacy choices

Your settings and saved markets stay available without analytics. With your permission, we use optional analytics to understand how the site is used.

Settings could not be loaded. Analytics remains off in this page.

Remembers your privacy choice and supports the features you request.

Measures public page visits and feature use. When enabled, a visitor cookie can link visits for up to 90 days. No advertising tracking.

You can change your choice here at any time. Declining does not limit access.