How Polymarket Markets Resolve: UMA, Disputes, Redemption

A Polymarket price is a live estimate of a probability, right up until it isn't. The Fed votes, the election gets called, the game ends - and the market has to turn that fact into a number every open position gets paid out against. Nobody at Polymarket flips that switch by hand: every market resolves through the same on-chain pipeline, whether the outcome is a coin flip or dead obvious the moment it happens. This blog has covered one resolution in the wild before; this post is the general mechanism, verified against Polymarket's own documentation and a market that resolved this month, as of 2026-09-25.

A bonded proposal, a 2-hour clock, and a fallback vote

Polymarket resolves markets with UMA's Optimistic Oracle - permissionless by design, so anyone can propose the winning outcome, not just Polymarket itself. A proposer picks the outcome and posts a bond, "typically $750 pUSD" per Polymarket's resolution docs. If nobody disputes it within a 2-hour challenge period, the proposal stands, the market resolves, and the proposer gets their bond back plus a reward. That is the entire mechanism for most markets: propose, wait two hours, done.

A dispute changes the shape of it. Disputing costs a matching counter-bond, and triggers a new proposal round; if that second proposal is also disputed, resolution escalates to UMA's DVM (Data Verification Mechanism), where UMA token holders vote on the correct outcome after a 24-48 hour debate window on UMA's Discord, with the vote itself taking roughly 48 hours. A disputed market therefore takes 4-6 days end to end against about 2 hours undisputed - and the bond design is what keeps the fast path honest: propose incorrectly, or before the event has actually concluded, and you lose the entire bond to whoever disputed you correctly. There is even a named outcome for a proposal that jumped the gun - "too early" - which pays the disputer exactly like a correct dispute would, and a rare "unknown/50-50" outcome for a genuinely ambiguous question, where every outcome token redeems for 50c instead of 0 or 100.

Even an obvious outcome runs the same clock, per market

Polymarket's "Fed Decision in September?" event resolved this month on the FOMC's 25bp hike, announced 2pm ET (18:00 UTC) September 16, 2026 - as uncontested an outcome as exists, confirmed within minutes by the Fed's own statement. Gamma's per-market data shows the pipeline ran anyway, and ran on its own clock for each Yes/No question inside the event rather than once for the event as a whole: the losing legs ("no change," both decrease contracts, and the 50+ bps increase) all carry closedTime 2026-09-16 20:02:43 UTC - almost exactly the ~2-hour undisputed window the docs describe, measured from the decision. The winning "increase 25 bps" leg didn't close until 21:40:27 UTC, nearly two hours later still. Same event, same real-world fact, five separate propose-and-resolve timers - because on UMA each binary question is its own request, not a single switch for the event.

RESOLUTION PATH                 DURATION
Propose (bond ~$750 pUSD)       instant
Challenge period                    2h
Debate period (if disputed)     24-48h
DVM token-holder vote (if any)    ~48h
undisputed total ~2h  ·  disputed total 4-6 days

The book stays open until the oracle actually finalizes

It is tempting to read a market's end date as when it stops trading - it isn't. Gamma's endDate is the expected resolution time, not a hard cutoff: acceptingOrders and closed only flip once a resolution actually finalizes through the pipeline above, so a market's book can keep trading, with real depth, well past the date printed on the question. What actually governs the outcome is each market's resolution rules - the source it resolves from, and how edge cases get called - which can differ in ways the headline question doesn't hint at. Polymarket can also publish an "Additional context" clarification after trading starts if a genuine edge case needs spelling out, though it cannot change what the question fundamentally asks. Reading those rules before you trade, not just the title, is worth the ten seconds: press i on any market page in polymarket-tui to open them.

A won position becomes a fixed $1 token, not a tick to sell into

Once a market's closed flag flips true, its outcomePrices freeze at ["1", "0"] and the CLOB book empties - a winning share stops being something you sell at a fluctuating price and becomes a token redeemable for a fixed amount. Redemption itself happens through the CTF collateral adapter on Polygon: it burns your ERC1155 outcome tokens through the CTF contract, releases the USDC.e collateral behind them, wraps it into pUSD, and returns it to your wallet - 100 winning tokens redeem for $100 pUSD, always, regardless of where the market last traded.

polymarket-tui doesn't perform that on-chain call itself - deposits, withdrawals, and redemption are on-chain operations the app leaves to the website. What it does do is track data-api's redeemable flag on every position, and flag a winner everywhere it shows up rather than let it look like a stale price:

POSITIONS
No change (Fed decision, Sept 2026)  won - redeem on web

The order panel carries the same warning if you try to sell a won position back into whatever book remains: "won - redeems at 100c on the web; selling here takes the bid" - a reminder that redemption always pays the full dollar, while a resting bid, if one still exists, might not. redeemable alone doesn't mean you won, either: data-api keeps returning losing shares as redeemable too, until you redeem them for nothing, and a genuine 50/50 resolution prices at 50c and still pays out - so the app checks redeemable together with price above 50c before it calls a position a win. Pressing o on a won position opens it on polymarket.com, where the actual redemption happens.

$ uv tool install polymarket-tui
$ polymarket-tui

No account is needed to browse markets, order books, and resolution rules. The install page has Homebrew and one-liner options.