Who Is Allowed to Ask for an Outcome Review?

Every prediction market sells the same promise: read the event correctly and you get paid. What almost no venue puts on the trading screen is the other half of it, which is what happens when the venue reads the event differently from you. That half decides whether being right is worth anything, and we have written about the gap before in Settlement Risk. The category now runs three answers to it at once, two of them under a single brand. This piece sets them side by side as of 21 November 2026, from the rulebooks rather than the marketing.

Two hours and a bond is the whole entry ticket on one side

On the international Polymarket, outcomes are not announced by the company. They are proposed to UMA, an optimistic oracle, and then they survive or they do not. Polymarket's help centre, in an article last updated on 11 January 2026, says that once a market is proposed for resolution it goes into a challenge period of two hours, and that the proposer posts a bond forfeited if the proposal fails. A companion article dated 5 April 2026 is blunter about who may object: anyone can dispute a proposed market resolution they believe was proposed in error, by posting a matching bond the page puts at roughly 750 dollars. That is a design statement, not a fee: standing to challenge an outcome is open to the whole world, and the price of using it is two hours of attention plus capital you forfeit if you are wrong.

The vote that decides it is weighted by how much UMA you staked

What a dispute buys is not a hearing in front of a neutral. It buys a ballot. UMA's documentation says a disputed request escalates to the Data Verification Mechanism, where holders who have staked the UMA token vote through a shielded commit and reveal scheme, and the answer is whichever option clears a consensus threshold those docs put at 65 percent of staked UMA. A fixed floor of 5 million staked UMA applies as a second quorum.

The incentives are explicit and, on their own terms, coherent. UMA's FAQ says voters who miss a round or vote against the majority lose 0.1 percent of their stake per vote, a correct disputer takes half the proposer's bond, and unstaking takes a week, so an attacker who bought enough stake to force a false answer would hold a depreciating token while that week ran. UMA puts a disputed request at two to four days; Polymarket's help centre describes a debate window of 24 to 48 hours and then a vote of about 48 hours. The deciding body, in short, is the people with the largest economic stake in the oracle, and the defence is that corrupting it should cost more than it pays.

A filing one day late turned a real sale into a No

The limit of that design showed in a market on whether Strategy would sell any bitcoin by 31 May 2026. The underlying fact is not in doubt, because the company reported it itself: in an 8-K signed 1 June 2026, Strategy disclosed a sale of 32 bitcoin between 26 and 31 May 2026. The sale happened inside the window, the filing landed one day outside it, and the market resolved No.

Two traders then did what the architecture does not contemplate. The Block, in a report by Danny Park on 7 July 2026, wrote that William Wood and Thomas Bush sued in the New York Supreme Court over the resolution, pleading breach of contract, breach of the implied covenant of good faith and fair dealing, unjust enrichment and New York consumer protection claims. The New York filing system refuses automated requests, so the pleadings are as reported by The Block, and we found no public decision by 21 November 2026 either. The point for a standard is not who should win. It is that the last appeal from an oracle that calls its outcomes immutable turned out to be an ordinary courtroom, by a route the product never mentions.

On the regulated side nobody outside the exchange can start the review

The US half of the brand is a designated contract market, and it answers the same question from the opposite direction. Rule 10.4 of the Polymarket US rulebook dated 30 September 2026 creates a Contract Outcome Review Process and opens by naming who may start one: the Chief Executive Officer, the Chief Compliance Officer, the Chief Operating Officer, the Head of Markets, or a designate, before settlement, at their sole discretion. A participant is not on that list. Rule 10.4(c) then says the company has full discretion in reviewing markets and that its determinations are final, and for contracts about a named natural person Rule 10.5(f) says determinations are not subject to review.

Contract terms repeat it. The terms for a 2026 gubernatorial margin contract dated 17 November 2026 make the exchange itself the Source Agency, list the official electoral authorities as primary sources and 23 named newsrooms as secondary ones, and end the settlement section by saying that once the exchange has determined an outcome, the determination is final and not subject to appeal or revision. The surface a trader actually reads compresses that into one line: the market resolution page in the Polymarket US documentation has a heading called Finality of Resolution and, under it, a single sentence pointing at the rulebook through a regulatory landing page rather than at the rule.

The same rulebook knows how to write a request window

Here is what turns that into a finding. The same rulebook does give participants a right to ask for a review, just not about the outcome. Rule 5.7(b) lets a participant request review of a trade, sets a deadline of fifteen minutes after execution, and defines a non-reviewable range of twenty cents around a calculated fair market value.

So the capability exists in the document. A request window, a deadline, a notification duty and a published test were all drafted once, for the question of whether a price was wrong, and not for the question of whether an outcome was wrong. Rule 5.7(b)(v) says price decisions are final and not subject to appeal either, so finality is the house style; what differs between the two cases is whether the person whose money is at stake can make the exchange look twice at all.

The other US exchange hands the same decision to two public directors

Because both venues sit under the same federal regime, the comparison is fair. Rule 7.1 of the KalshiEX rulebook, version 1.32, also puts initiation in the exchange's sole discretion, so a trader cannot start that one either. Three things downstream of it differ, and all three are measurable.

Who decides: Kalshi Rule 2.7(e) makes it a standing Outcome Review Committee of three members appointed by the Regulatory Oversight Committee, at least two of whom must be Public Directors, while Polymarket US Rule 10.4(a) makes it four named executive officers or a designate. The clock: Kalshi Rule 7.1(b) requires a determination within 24 hours of initiation, while Polymarket US sets no outer limit. And when a holder of an open position learns anything: Kalshi Rule 7.1(a) says the exchange will post on its website that an outcome is under review, while Polymarket US Rule 10.4(b) commits to notice of the final determination. Both say the result is final. Only one lets you know the question is open while it is open.

What the federal floor actually asks for is a facility

What the regulations actually require is less than readers assume. Core Principle 14 for designated contract markets, at 17 CFR 38.750, says the board of trade shall establish and enforce rules regarding, and provide facilities for alternative dispute resolution as appropriate for, market participants and any market intermediaries. That is a duty to provide a forum, not a duty to let anyone appeal a settlement.

The texture is in the guidance. Appendix B to Part 38 spells out an objective and impartial decisionmaker, adequate notice, an opportunity to be heard and a prompt hearing, and twice says the procedures should be available on a voluntary basis to customers who are not eligible contract participants. Set that against Chapter 9 of the Polymarket US rulebook, where claims among participants and claims against the exchange shall be resolved through binding arbitration in New York, and whose Rule 9.1(e) makes failing to arbitrate, or filing in court first, itself a rule violation exposing the participant to disciplinary proceedings. Kalshi writes the same chapter differently. Its Rule 10.1(a) begins with the words if so elected by a Participant.

An Individual Participant is defined as someone who is not a Customer

There is a reason the word customer keeps appearing in the regulations and not in the architecture. 17 CFR 166.5 protects a customer who signs an arbitration clause before any dispute exists: it says use of dispute settlement procedures shall be voluntary, forbids making the clause a condition of using the service, preserves a reparations election within 45 days, and requires a list of qualified forums including one where a majority of decisionmakers are not associated with the exchange.

Two textual facts decide whether any of that reaches a retail trader here. The rule binds Commission registrants, a list naming brokers, merchants, pool operators and advisers, and not a contract market. And its definition of customer in 166.5(a)(2) excludes a participant in the designated contract market. Meanwhile the Polymarket US rulebook defines an Individual Participant as a participant who is a natural person and who is not a Customer. Our own count over the full 35,451 words of that rulebook: voluntary zero times, reparations zero times, 166.5 zero times, Core Principle 14 zero times, appeal 81 times and nearly all of it disciplinary. This is not an accusation of a breach, and it is not legal advice. It describes how a retail account can end up outside a protection written for retail accounts, by being onboarded as a participant instead of a customer.

What the statute leaves is narrow. The federal private action against an exchange is 7 U.S.C. section 25(b), which makes a registered entity liable for actual damages when it fails to enforce rules it is required to enforce, but subsection (b)(4) requires the person suing to establish that the entity acted in bad faith, and (b)(5) makes that the exclusive remedy under the Act. So on the oracle side the decision is made by stakeholders voting their own money and the backstop is state contract and consumer law, while on the exchange side the decision is made by officers under a federal regulator and the backstop requires proving bad faith, in a forum the rulebook reaches only after mandatory arbitration. Neither offers what a reader pictures on hearing the word appeal.

The choice that shapes your recourse is made in the first thirty days

The procedures for a claim against Polymarket US are not in the rulebook at all. Rule 9.2(b) sends them to the website terms of use and adds that the company reserves the right to modify those terms from time to time. The terms of use dated 27 May 2026 set out an informal process as a condition precedent with a tolling period of 60 business days, then arbitration before a sole arbitrator under the JAMS Comprehensive Arbitration Rules, on an individual basis only, with a class action waiver and a small claims carve out.

They also contain the clause a reader most needs to see early. Arbitration can be declined, but only by written notice to the company's disputes address received no later than 30 days after entering the agreement, and a successful opt out consents to the exclusive jurisdiction of the courts of New York County. That window opens at sign up and closes long before most people hold a position large enough to care about, or have seen how the venue behaves when an outcome is contested. That is the clearest Directive 04 problem in the comparison: the most consequential decision about recourse is put to the user at the moment they understand the product least.

Five things either model could publish without changing its design

None of this asks a venue to give up finality. Finality is defensible, and open ended settlement is worse than a decision that sticks.

  • A recourse box on the market page. Who can ask for this outcome to be looked at, by when, decided by whom, published where. Four lines next to the rules, not a link to an 84 page PDF. Directive 02 asks for settlement sources in plain language on the surface where the trade happens, and the recourse path belongs there too.
  • Notice at the start of a review, not only at the end. Kalshi Rule 7.1(a) does this, and a holder who learns an outcome is under review can at least stop adding to the position.
  • A clock. Kalshi's 24 hours in Rule 7.1(b) is a published commitment. A review with no outer limit is indistinguishable, from outside, from one that is not happening.
  • A standing decider with a majority from outside the business. Two Public Directors out of three is a written answer to the conflict a venue has when it determines outcomes on its own book.
  • The opt out explained when it matters. A 30 day arbitration opt out mentioned once at sign up, inside a 17 page terms document, is technically disclosed and practically invisible. Repeat it at the first withdrawal.
  • A roll call, for the oracle model specifically. If the deciding body is weighted by stake, publishing which addresses voted and with what weight on each disputed market is what makes the claim of an honest majority checkable.

Finality is a design choice and it should be sold as one

Nothing here says the venues are hiding anything. All of it is published, and the UMA documentation is more candid about its own weaknesses than most financial disclosure. The problem is sequencing, the same one we found in the brand's two rulebooks when we looked at which regime holds a given position and at what a clause can redefine after the first day of trading. A trader meets the resolution rules at the moment of the trade and the recourse rules at the moment of the loss, and the second set is longer, harder to find, and in one case kept in a document the venue may change. Directive 01 commits a signatory to publishing resolution rules before a market opens and not revising them while money is at risk. The route by which a resolution can be questioned is part of those rules. Treat it that way, put it on the card next to the price, and the word final starts to mean something a reader agreed to rather than something they discover.

Share X LinkedIn Email