# Prediction Market Resolution Disputes in 2026: Which Oracle Holds Up

> In 2026, UMA token holders settled Polymarket disputes worth hundreds of millions while many voters held positions, and outcomes are final once voted. Kalshi made settlement errors too, but reversed or compensated them and answers in court. Our view is that rules locked at listing, settled by an accountable operator, hold up best.

Published 2026-11-11 by PM Crypto Notes. Canonical: https://pmcrypto.blog/oracle-problem-2026/

## Key points

- More than 1,150 Polymarket markets went to dispute in 2026 by mid-May, already past the 2025 total, per a WSJ analysis reported by Finance Magnates.
- The WSJ found at least 60% of active UMA voters linked to Polymarket accounts, and a voter with a stake in about one in five disputes.
- UMA's market cap was $28.1 million on November 11, 2026; Polymarket's Iran peace-deal markets alone drew $478.91 million, per Calcalist.
- Kalshi reversed its January 2026 NFL misgrades by the next morning and reimbursed fees and net losses on the Khamenei market.
- The CFTC's June 10, 2026 proposal names settlement on 'clear, objective and publicly verifiable criteria' as an integrity factor.


Our position is that token-holder voting did not hold up as the final word on disputed prediction markets in 2026, and that the design which does hold up is the one Polymarket already runs for US users: rules fixed at listing, a named source, and an operator that answers to a regulator and a court, with the optimistic proposal kept as a fast path for the markets nobody contests.

## How does Polymarket's oracle settle a market?

Polymarket's international venue settles through UMA's optimistic oracle. A proposer posts an outcome and a bond; per [Polymarket's help center](https://help.polymarket.com/en/articles/13364551-how-are-markets-disputed), anyone can challenge within a two-hour window by posting a matching bond, "usually $750." A disputed question goes to UMA token holders, a vote of about 48 hours. [UMA's documentation](https://docs.uma.xyz/protocol-overview/how-does-umas-oracle-work) describes a 24-hour commit and 24-hour reveal, resolution once 65% of staked UMA backs one answer, and slashing of stakers who skip the vote or land in the minority. Polymarket's help center adds: "Once finalized by UMA, outcomes are immutable."

Since August 2025, under UMA's MOOV2 upgrade, only whitelisted addresses can propose; [The Block reported](https://www.theblock.co/post/366507/polymarket-uma-oracle-update) an initial list of 37. Anyone can still dispute.

Our own pull of Polymarket's gamma API at 05:07 UTC on November 11, 2026 shows how much rides on small bonds. Across 24,013 open contracts in the 300 deepest events, the API's `umaBond` field read $250 on 17,520 and $500 on 5,928. The Thunder vs. Trail Blazers moneyline, which traded $2.0 million in the prior 24 hours, carried a $250 bond figure and a custom challenge window of 900 seconds. Two of the 24,013 contracts showed a disputed status: whether Lebanon, and whether Venezuela, will recognize Israel by December 31.

## What went wrong in 2026?

The undisputed path works. Polymarket has said about 0.2% of contracts trigger a UMA vote, [according to Yellow](https://yellow.com/news/polymarket-7m-ukraine-israel-bet). The problem is the disputed path, which is where the money is.

A Wall Street Journal analysis, [reported by Finance Magnates](https://financemagnates.com/fintech/polymarkets-arbitration-model-faces-conflict-of-interest-questions) on May 18, 2026, found more than 1,150 Polymarket markets had gone to arbitration in 2026, already past the 2025 total. At least 60% of active UMA voters were linked to Polymarket accounts. In about one in five disputes, at least one voter held a stake in the outcome, and in most disputes the 10 largest wallets held more than half the voting power. One voter, Scout, told the Journal: "You can either have traders with a conflict of interest, or morons with no conflict of interest." Polymarket founder Shayne Coplan called the process "messy."

Three cases show the pattern.

**Strategy's bitcoin sale.** The market asked whether Strategy would sell any bitcoin by May 31. On June 1, Strategy's 8-K disclosed a sale of 32 BTC between May 26 and May 31. Polymarket then posted context stating, [per The Block](https://theblock.co/post/403600/polymarket-upholds-no-outcome-strategy-bitcoin-sale-market), "Confirmation achieved outside of the market's time frame does not qualify." After two disputed proposals, a UMA vote with about 98.6% of voting power resolved No on June 3. [Galaxy Research](https://www.galaxy.com/insights/research/strategy-bitcoin-sale-polymarket-resolution-dispute-may-2026), which disclosed a financial interest in YES shares, wrote: "Everyone who bought YES predicted the future correctly, and the market just told them they were wrong." Two traders [sued Polymarket](https://www.theblock.co/post/407368) in New York Supreme Court on July 3.

**The Iran peace deal.** A Polymarket contract asked whether the US and Iran would agree to a "permanent peace deal" by June 15. [The Block reported](https://www.theblock.co/post/405117) more than $120 million on that contract, and more than 99% of voting power backing Yes while the vote was still open. On June 17, Trump said of the agreement: "It's a memorandum of understanding." [Calcalist reported](https://www.calcalistech.com/ctechnews/article/wgc50wv09) that the vote resolved Yes, put total volume across the deadlines at $478.91 million, and cited one critic's count that the top seven holders, all voting Yes, had 50.3% of voting power. That count comes from a losing trader and has not been independently verified.

**The Hezbollah ceasefire.** [Yellow reported](https://yellow.com/news/polymarket-7m-ukraine-israel-bet) that 87% of voting UMA holders ruled that a truce between Israel and the Lebanese government counted as the deal a market asked about. A trader who had staked $567 against a deal complained to token holders and got nowhere.

We do not claim any of these outcomes was wrong. Our point is narrower: in each, the question turned on reading contract language, and the readers were anonymous, concentrated, often positioned, and final.

## Does UMA's security math cover markets this size?

UMA's documentation frames security as a cost of corruption: if 65% of stake voted against reality, UMA "would likely depreciate sharply," punishing the attackers. On November 11, 2026, CoinGecko put UMA's entire market capitalization at $28.1 million, at $0.306 a token. Calcalist's figure for the Iran markets alone is $478.91 million. Staked UMA is some fraction of that $28.1 million, and the 65% needed to decide a vote is a fraction of the stake. Our read is that once a single question is worth many times the token, the deterrent the design relies on no longer scales with what it protects.

## How does Kalshi's centralized resolution compare?

Kalshi has had bad settlements too. On its "Ali Khamenei out as Supreme Leader?" market, with more than $50 million in volume, Kalshi's CFTC-filed terms settled a death at the last traded price before death, which CEO Tarek Mansour said was recorded at 1:14 AM ET. [The Block reported](https://www.theblock.co/post/391678/kalshi-ceo-defends-khamenei-market-design-after-backlash-says-platform-will-reimburse-all-fees) on March 1, 2026 that the filed terms and the market page used different wording, which Kalshi called "grammatically ambiguous." A proposed [class action](https://truthinadvertising.org/wp-content/uploads/2026/03/Golan-v-Kalshi-complaint.pdf), Golan v. KalshiEX, was filed in the Southern District of New York on March 24; the plaintiff says his YES contracts settled at $0.29. In January 2026, [OddsShopper reported](https://www.oddsshopper.com/articles/prediction-markets/kalshi-stat-corrections-settlement), Kalshi misgraded several NFL win-total markets at once.

What happened next is the difference. Kalshi first offered refunds of cost, then by the next morning paid correct YES holders "their full contract settlement value of $1." On Khamenei, it reimbursed all fees and, [per Finance Magnates](https://financemagnates.com/fintech/khamenei-out-market-becomes-legal-headache-for-kalshi), covered net losses, which Mansour said "resulted in a financial loss for the firm." OddsShopper reports Kalshi has since adopted a policy of refunding trading fees on disputed settlements in proportion to traders' losses. Under Rule 7.1, Kalshi's Outcome Review Committee must decide a reviewed outcome within 24 hours.

Polymarket has made the same choice for its own regulated venue. [DeFi Rate's explainer](https://defirate.com/prediction-markets/how-contracts-settle/) says Polymarket US settles centrally through Polymarket Clearing and "does not rely on blockchain or UMA oracle voting."

Regulators are moving the same way. The CFTC's June 10, 2026 proposed rule on event contracts, [summarized by Skadden](https://www.skadden.com/insights/publications/2026/06/cftc-issues-notice-of-proposed-rulemaking), lists whether a contract "settles based on clear, objective and publicly verifiable criteria" among its integrity factors.

## What is the strongest case against our view?

The strongest counter-argument is that the oracle was never the problem. The Strategy dispute began with Polymarket's own clarification, and Kalshi's Khamenei dispute began with its own ambiguous wording. Centralization also concentrates conflicts: Mansour reposted a Kalshi post promoting the Khamenei odds, and DeFi Rate notes that Kalshi has no independent appeal and quotes the Massachusetts attorney general's complaint that its process "does not serve as an independent intermediary." UMA, by contrast, records every vote on-chain.

We agree that bad wording caused most of 2026's fights on both venues, and that Kalshi lacks an outside appeal. But the two designs fail differently. When Kalshi gets it wrong, an identifiable company reverses, pays, or gets sued in federal court. When UMA voters get it wrong, the outcome is immutable, the voters are anonymous, and the traders who sued over Strategy named Polymarket and its executives anyway. Accountability ends up with the operator regardless; the centralized design admits it and gives it a remedy.

The design we think holds up: criteria locked at listing, with no clarifications that change the standard after money is in; a named primary source; operator settlement with published reasons; and an appeal outside the operator. Optimistic proposals remain a good fast path for the undisputed majority.

## What would change our mind

- A UMA change that bars voters from holding positions in the markets they decide, with evidence it is enforced on-chain.
- Data showing UMA-decided disputes were overturned or regretted less often than Kalshi Outcome Review decisions over a comparable set of markets.
- Staked UMA rising to a value comfortably above the largest single disputed market.
- Kalshi or Polymarket US making a settlement error on a large market and refusing to reverse or compensate it.


## FAQ

### How does Polymarket resolve disputed markets?

A whitelisted proposer posts an outcome with a bond, usually $750, and anyone can challenge it within a two-hour window. Disputed questions go to a vote of staked UMA token holders, which takes about 48 hours, and Polymarket's help center says outcomes finalized by UMA are immutable.

### How often are Polymarket markets disputed?

Polymarket has said about 0.2% of contracts trigger a UMA vote. A Wall Street Journal analysis reported in May 2026 counted more than 1,150 disputed markets in 2026 to that point, already more than in all of 2025.

### How does Kalshi resolve markets?

Kalshi's markets team settles contracts against sources named in its CFTC-filed terms. Under Rule 7.1, an Outcome Review Committee can review a disputed outcome and must decide within 24 hours, and its determinations are final. DeFi Rate notes Kalshi has no independent appeal mechanism.

### What happened with the Polymarket Strategy bitcoin market?

Strategy disclosed on June 1, 2026 that it sold 32 BTC between May 26 and May 31. Polymarket added a clarification that confirmation outside the time frame did not qualify, and a UMA vote with about 98.6% of voting power resolved the market No. Two traders sued Polymarket on July 3.

### Does Polymarket US use the UMA oracle?

No. According to DeFi Rate, Polymarket US settles centrally through Polymarket Clearing under its CFTC rulebook and does not rely on UMA oracle voting. The international Polymarket venue still resolves through UMA.

## Sources

- [How are markets disputed?](https://help.polymarket.com/en/articles/13364551-how-are-markets-disputed), Polymarket Help Center
- [How does UMA's Oracle work?](https://docs.uma.xyz/protocol-overview/how-does-umas-oracle-work), UMA Documentation
- [UMA's oracle update to limit Polymarket resolution proposals to whitelisted parties to improve market outcomes](https://www.theblock.co/post/366507/polymarket-uma-oracle-update), The Block, 2025-08-12
- [Polymarket's Arbitration Model Faces Conflict-of-Interest Questions](https://financemagnates.com/fintech/polymarkets-arbitration-model-faces-conflict-of-interest-questions), Finance Magnates, 2026-05-18
- [Polymarket's $7M Ukraine Ghost Returns As Israel Bet Sparks Outrage](https://yellow.com/news/polymarket-7m-ukraine-israel-bet), Yellow, 2026-05-18
- [Strategy Sold Bitcoin in May. Polymarket Says It Didn't.](https://www.galaxy.com/insights/research/strategy-bitcoin-sale-polymarket-resolution-dispute-may-2026), Galaxy Research, 2026-06-04
- [Polymarket's UMA vote upholds 'No' outcome on Strategy bitcoin sale market despite backlash](https://theblock.co/post/403600/polymarket-upholds-no-outcome-strategy-bitcoin-sale-market), The Block, 2026-06-04
- [Two traders sue Polymarket over disputed resolution of Strategy bitcoin sale market](https://www.theblock.co/post/407368), The Block, 2026-07-07
- [Trump comments fuel dispute over Polymarket's $120 million 'permanent' Iran peace deal market](https://www.theblock.co/post/405117), The Block, 2026-06-17
- [Polymarket's $354M Iran Dispute Exposes Prediction Markets' Settlement Problem](https://financemagnates.com/fintech/polymarkets-354m-iran-dispute-exposes-prediction-markets-settlement-problem), Finance Magnates, 2026-06-16
- [Was the Iran deal real? Inside the Polymarket decision traders are fighting](https://www.calcalistech.com/ctechnews/article/wgc50wv09), Calcalist (CTech), 2026-07-06
- [Kalshi CEO defends Khamenei market design after backlash, says platform will reimburse all fees](https://www.theblock.co/post/391678/kalshi-ceo-defends-khamenei-market-design-after-backlash-says-platform-will-reimburse-all-fees), The Block, 2026-03-01
- ['Khamenei Out' Market Becomes Legal Headache for Kalshi](https://financemagnates.com/fintech/khamenei-out-market-becomes-legal-headache-for-kalshi), Finance Magnates, 2026-03-04
- [Golan v. KalshiEX LLC et al., Class Action Complaint, No. 1:26-cv-02394 (S.D.N.Y.)](https://truthinadvertising.org/wp-content/uploads/2026/03/Golan-v-Kalshi-complaint.pdf), Truth in Advertising (court filing), 2026-03-24
- [Kalshi Bet Marked Lost After Team Won? The NFL 9-Win Case](https://www.oddsshopper.com/articles/prediction-markets/kalshi-stat-corrections-settlement), OddsShopper, 2026-08-14
- [How Contracts and Dispute Resolution Works at Prediction Markets](https://defirate.com/prediction-markets/how-contracts-settle/), DeFi Rate, 2026-09-28
- [CFTC Issues Notice of Proposed Rulemaking on Prediction Markets](https://www.skadden.com/insights/publications/2026/06/cftc-issues-notice-of-proposed-rulemaking), Skadden, 2026-06-17
- [UMA price and market cap (pulled November 11, 2026, 05:13 UTC)](https://api.coingecko.com/api/v3/simple/price?ids=uma&vs_currencies=usd&include_market_cap=true), CoinGecko API, 2026-11-11
- [Gamma API: events endpoint (300 events by liquidity, pulled November 11, 2026, 05:07 UTC)](https://gamma-api.polymarket.com/events?active=true&closed=false&order=liquidity&ascending=false&limit=100&offset=0), Polymarket, 2026-11-11