The request timing, fallback settlement time, evidence rules, and fee-policy version were fixed before the first bet.
Request eligibleToken weighted arbitration, like Kleros or UMA with Polymarket, is not incentive aligned with traders, neither are centralized arbitration services like Kalshi. See how in the case studies in the articles below:
Ukraine mineral deal clearest whale case
Reporting traced a disputed Yes resolution to one UMA whale voting through three accounts with about five million UMA, roughly a quarter of the vote. That is the cleanest example of token weight pushing the arbiter itself.
Voters with positions systemic evidence
WSJ and Barron's reported recurring overlap between UMA voting power and Polymarket exposure, including large-wallet concentration and voters resolving markets they could also bet on.
Zelenskyy suit public scandal
Photos and major-media descriptions were not enough to stop a No resolution after the UMA dispute path. The word "suit" left room for argument, but the public lesson was brutal: common-sense evidence still had to survive a token vote.
Strategy Bitcoin sale plain reality lost
Strategy disclosed that it sold Bitcoin during the market window, yet the corresponding Polymarket contract resolved No after the dispute process. This is the best example of a correct event prediction losing to resolution mechanics.
Kalshi's Khamenei carveout platform as judge
Kalshi shows the centralized version of the same failure: the house can write the escape hatch and then judge it. Plaintiffs alleged Kalshi used its Khamenei death carveout to avoid full Yes payouts and settle at the last traded price. The venue benefits from limiting payout, precedent, and reputational exposure; users want the event paid as written. DemoThemis removes that conflict. Rules are fixed before escrow, and hard cases go to randomly drawn, conflict-screened jurors outside the venue's balance sheet.
Kalshi's post-event listing house sets the clock
How Kalshi creates a market. The venue lists every contract itself. It self-certifies a contract template with the CFTC, then fills that template in to open individual markets on a theme whenever it decides to. No trader can list a contract, and no trader sets the opening time. The venue picks the question, the wording, and the moment the book opens — and then settles what it opened.
What that allowed. On 11 August 2026 a trader documented an earthquake market that Kalshi opened about an hour after the earthquake had already happened; seven markets from that template resolved about two hours later. Nothing was being predicted. The answer already existed, and the only edge was having read the news. Whoever took the other side was paying into a question that was already closed, and the venue earned its fee either way.
Why OmenMarketMaker has no such lever. There is no listing desk whose clock anyone can set: a market opens because someone bet on it, so the venue never holds a private moment to time. Anyone can still open a market on something already decided — but it cannot become a trap. Every term is fixed at creation and readable before a single side is funded, and a one-sided pool goes nowhere, because graduation needs both sides to hold backing and funded two-sided depth for 24 continuous hours, and nobody funds the side that has already lost. The house does not rule on it either: a final YES or NO comes from drawn jurors, not from the venue that opened the book.
The conclusion: a prediction market needs an unbuyable arbiter, and a book no one can time. UMA shows the whale failure. Kalshi's Khamenei dispute shows the house failure at settlement, and its post-event earthquake listing shows the same house control one step earlier, at the moment the book opens. OmenMarketMaker removes the venue from both ends: it does not choose what gets listed, and it does not decide the final verdict.
Arbitration
How OmenMarketMaker Resolutions Are Superior To All Other Prediction Markets For Traders:
How a market ends with an unbuyable judge
The event happens: No new YES or NO tokens can be minted after the market closes. Existing tokens can still trade until finality while their collateral remains in escrow and the five-step resolution path begins.
-
The market closes under fixed rules
The question, permitted answers, evidence rules and sources, the evidence window, the request timing, the court route, the fee-policy version, and the repayment and fallback settlement rules were all set before trading. No operator can rewrite any of them after seeing the money or likely outcome.
-
The evidence record freezes
During the fixed window, anyone can add valid material from a permitted source. When it closes, OmenMarketMaker freezes the complete record so nobody can add, remove, or reorder evidence for that case.
-
Anyone can chip in to pay the court
When the market closes, its page shows a funding panel: the exact fee, how much is still missing, a countdown, and every wallet that has paid in beside the amount it is owed back. Anyone adds any amount and can withdraw it while the countdown runs. The contribution that completes the total locks the whole fee inside DemoThemis — OmenMarketMaker never holds it — submits the frozen case and evidence record, and opens one case in the same transaction. If the countdown ends first, or opening fails, every contribution returns and the pool is unchanged. Contributors supply only the market ID and money; they choose nothing about the case. The same panel funds an appeal later.
-
Verified humans decide the case
DemoThemis accepts the case before a public draw, seats verified humans, and aggregates private ballots. A new appeal panel is drawn only when contributors choose to fund it.
-
The signed ruling returns
After a final YES or NO, contributors are paid back first — each exactly what they put in, taken pro rata from both sides' backing — before the Omen fee and winning-token payouts. An insufficient-information ruling can be appealed like any other; only when it still stands after every funded appeal does the court keep the bond. Then, after the fixed cooldown, a new funding panel opens for a fresh attempt before the final fallback time. New YES or NO tokens remain closed to minting, and every side's original backing stays intact. Once the final fallback time is reached, the market voids instead.
The service boundary stays deliberately narrow: the locked case details and crowd-funded court fee enter DemoThemis; a signed ruling comes back. Outcome collateral, the reimbursement rule, market fee, and settlement stay with OmenMarketMaker.
The crowd's bond pays the court fee calculated by fixed rules before random juror selection tied to that case, private ballots, and panel consensus.
Signed final rulingYES or NO pays contributors back and pays winning tokens from the remaining pool. Insufficient information pays nobody back and leaves the market backing untouched.
Precommitted settlementEvery accepted resolution request uses the court: verified humans drawn by a public source of unpredictable numbers and voting in encrypted ballots. Contributors pay the initial jury up front; another panel is drawn only if contributors fund an appeal. Reimbursement waits until every appeal is final, so a provisional ruling cannot determine who pays. If the exact court fee is at least the pool, no request can promise a full directional reimbursement and the set fallback settlement rule applies instead. Every completed court case contributes to the shared juror-quality record, and independently checkable outcomes can add later confirmation.
Endless appeals are what make it robust
Every other arbitration system has to be right the first time. This one does not, and that is what makes it the most robust arbitration a prediction market can have. Any ruling can be appealed, the appeal is crowd-funded from the market page in a few taps, and every appeal draws a fresh panel larger than the one before it. Two things turn that ladder into robustness. The first is the incentive. Funding an appeal costs money and a side that turns out to be wrong loses what it paid, so only real conviction buys another round, and the side that is actually right gains ground every time it pays. The second is what this does to the accuracy a single juror needs. A panel whose answer is final has to be excellent. A panel that can be appealed does not, because a mediocre bench still lands on the right final answer when being wrong is cheap to correct and expensive to defend. The interface carries as much weight as the mechanism. Appealing is a funding panel on the market page with a countdown, any amount, withdrawable while it runs. It is not a governance proposal, not a token vote, not a support ticket. Anyone who thinks the court got it wrong can put money behind that opinion in seconds, which is the whole reason the ladder ever gets climbed.
The game below is that argument, played out on a bench nobody would call good. Set how often a single juror is right, watch each side work out whether another panel is worth what it costs, and watch the meter: every appeal that gets funded draws a larger panel and raises the chance the final answer is the true one, while the side that turns out to be wrong pays for every round it asked for.