Resolution Risk: How Prediction Markets Decide Who Won (and When It Goes Wrong)

When a prediction market pays out, a specific machine decides who won. On Kalshi, the machine is the exchange itself: outcomes are determined in-house under contract rules filed with the CFTC, and an internal committee has the final word. On Polymarket’s global platform, it is a crowd with money at stake: a whitelisted proposer posts a bond to submit the outcome, anyone can dispute it, and a twice-contested market ends in a vote of UMA token holders. Resolution risk is the gap either machine leaves between what you watched happen and what the rules actually pay — and it has decided tens of millions of dollars in ways traders did not see coming. Below: both systems as of August 2026, three famous failures, and the habits that keep you off the wrong side of the fine print.

How does Kalshi resolve a market?

Kalshi’s model is centralized and regulated. Every contract carries terms filed with the CFTC that spell out the settlement rules and name the source agency whose data decides the outcome — league statistics for sports, government releases for economic prints, official announcements for news. When a market expires, Kalshi’s own markets team determines the outcome from that source, usually within hours. Traders can flag a market through the request-to-settle tool, but it functions as a suggestion; the decision belongs to the exchange.

Two rulebook provisions matter when things get contested. A Market Outcome Review hands the call to an internal Outcome Review Committee that must rule within 24 hours, and its determination is final. And under Rule 6.3(c), if an outcome is deemed unresolvable, Kalshi can settle every position at the last traded price instead of $1 or $0. There is no external arbitrator and no appeal inside the platform; the backstops are the CFTC and the courts. Our Kalshi review scores how that structure performs.

The federal backstop is not theoretical. In July 2026, after a Michigan state court ordered Kalshi to void Michigan users’ trades in the state attorney general’s suit, the CFTC stayed the exchange’s plan to unwind the positions and directed it to honor the trades, reasoning that retroactively unwinding executed contracts risks distorting the whole market. The last word on an executed trade belongs to the federal regulator — not the exchange, and not a state court.

How does Polymarket resolve a market?

Polymarket’s global platform outsources truth to UMA’s optimistic oracle — a system that assumes a proposed answer is correct unless someone pays to disagree. The flow, per Polymarket’s documentation: once an outcome is known, a proposer submits the result and posts a bond, typically $750. A two-hour challenge window opens; if nobody disputes, the market resolves — the path nearly all markets take. A dispute (which requires matching the bond) triggers a fresh proposal round, and a second dispute escalates to UMA’s token-holder vote: about two days of evidence and debate, then a two-day vote in which more tokens means more weight. A fully disputed market takes four to six days; the vote’s losing side forfeits half its bond to the winner; rare unresolvable cases settle 50/50, every share redeeming at 50¢.

After years of premature and sloppy proposals, UMA governance approved a “managed” oracle for Polymarket in August 2025: proposing is now restricted to a whitelist of proven proposers — initially 37 addresses: Risk Labs and Polymarket staff plus accounts with 20-plus proposals at 95%-plus accuracy — while disputing remains open to anyone. Note what the reform did not touch: a twice-disputed market still ends in a token-weighted vote.

One more distinction, as of August 2026: Polymarket US, the CFTC-designated exchange the company operates for American customers, resolves its markets in-house under filed rules, Kalshi-style; UMA governs the global platform. Same brand, two judges — our Polymarket review breaks down which one your trade lives on.

The rules decide who gets paid — not the event, and not your reading of the news.

Three failures worth studying

The Ukraine-minerals vote (March 2025). A $7 million market asked whether Ukraine would agree to a minerals deal with the Trump administration before April. No deal had been announced, but in late March the price snapped from single digits to 100% and the market resolved yes — after a holder controlling roughly five million UMA tokens across three accounts cast about a quarter of the dispute vote, per The Block. Polymarket called it an unprecedented governance attack, said it was not a market failure, and issued no refunds. It remains the cleanest demonstration of the oracle’s weakness: enough tokens can outvote reality.

The Strategy bitcoin sale (June 2026). A market with more than $50 million in volume asked whether Strategy, Michael Saylor’s bitcoin-treasury company, would sell any bitcoin by May 31, 2026. An SEC filing published June 1 disclosed a sale of 32 BTC between May 26 and 31 — inside the window, disclosed a day after it closed. Two proposed “no” resolutions were disputed, the question went to a token vote, and Polymarket posted a clarification telling voters that “confirmation achieved outside of the market’s timeframe does not qualify.” The market resolved no. In July, two traders sued Polymarket and its CEO in New York state court, alleging the confirmation requirement was added after the fact; the complaint seeks the full $1-per-share payout and is pending as of August 2026. The lesson is mechanical: when wording meets an inconvenient fact pattern — sold by May 31, filed June 1 — the process, not the event, decides the money.

Kalshi’s Khamenei carveout (February–March 2026). After Iran’s supreme leader Ali Khamenei was killed in U.S.–Israeli airstrikes in late February 2026, Kalshi’s market on whether he would be out as supreme leader — more than $54 million in trades, per The Hill — did not pay yes holders $1. The rules contained a death carveout: if the leader left office by dying, the market settles at the last traded price before the death. CEO Tarek Mansour said Kalshi deliberately writes rules so no market settles directly on a death, which he said U.S.-regulated entities may not offer; Kalshi reimbursed fees and refunded post-death entries, but paid yes holders the pre-death price. A proposed class action filed in March 2026 alleges Kalshi refused to honor winning contracts; Kalshi’s answer is that the carveout was printed in the rules all along. Either way: the fine print most traders never opened was the entire ballgame.

What did the Wall Street Journal find about the judges?

In May 2026, a Wall Street Journal investigation of Polymarket’s dispute system reported that in most disputed markets, more than half the UMA votes came from the ten largest wallets; at least 60% of active UMA voters could be linked to Polymarket betting accounts; and roughly one in five disputes included a voter with a financial stake in the market they were ruling on. Nothing prevents a voter from judging a market they hold a position in, and most voters are anonymous. By mid-2026, reporting put the year’s disputed-market count past 1,150, already ahead of all of 2025. The oracle has settled thousands of markets without incident, but the structural fact stands: the people deciding contested payouts can be the people holding the positions.

Five habits that protect you

Read the resolution criteria, not the title. The title is marketing; the rules are the contract. The death carveout and the timeframe language were in the rules before they mattered. Even Polymarket’s own documentation tells traders to read the rules before trading; only the rules define the answer.

Check the resolution source before you buy. Kalshi names a source agency in each contract; Polymarket names sources in the market description. A market that settles on one official document is a different asset from one that settles on a consensus of credible reporting; if you cannot tell which document decides it, that is your answer.

Know the dispute clock. On Polymarket, the challenge window after a proposal is two hours; a wrong proposal that slips through unchallenged is final. On Kalshi, raise concerns while settlement is pending, via the request-to-settle tool; once the review committee rules, it is final. In both systems, act before settlement, not after.

Compare venues on the same question. At the February 2026 Super Bowl, both platforms ran markets on whether Cardi B would perform at halftime. She appeared during Bad Bunny’s set; whether that counted as performing split the venues. Kalshi, citing ambiguity, settled at the last traded price — 26¢ for yes — under Rule 6.3(c), while Polymarket’s differently worded market resolved yes at $1. Same stage, opposite payouts — both correct under their own rules.

Price ambiguity like the cost it is. A market whose wording could plausibly split — sells versus confirms, performs versus appears, deal versus announced deal — carries a risk the odds never show. Our glossary keeps working definitions of resolution risk and the rest of the vocabulary these disputes turn on.

Quick answers

How long does Polymarket take to resolve a market? Undisputed: about two hours after an outcome is proposed. Disputed twice: four to six days including debate and the UMA vote, per Polymarket’s documentation as of August 2026.

Can a wrongly resolved market be reversed? Effectively no. A concluded UMA vote is final, and Polymarket declined refunds even in the case it called a governance attack. Kalshi’s review committee is final under its rulebook, though it has voluntarily reimbursed fees in contested markets. The outside avenues — CFTC complaints and lawsuits — are slow; the suits above are pending as of August 2026.

Do Kalshi and Polymarket ever resolve the same event differently? Yes. The Cardi B halftime markets settled in opposite directions in February 2026 because the platforms wrote different rules for the same moment. Cross-platform price gaps sometimes reflect exactly this: the contracts are not the same bet.

Who has the final say on a disputed market? On Kalshi: its Outcome Review Committee, under CFTC oversight — and in July 2026 the regulator stepped in to protect executed trades. On Polymarket’s global platform: the UMA vote, which Polymarket cannot override. On Polymarket US: the exchange itself, under CFTC-filed rules.