One Polymarket disputer is 80 for 80. Four others lost $23,000 in a single night. The $500 bond decides who gets paid.
Every challenge to a Polymarket result starts with a $500 bond. We followed 470 of them from the Polygon dispute to the UMA vote that settled it, and found that the winners and losers are the people posting bonds, not the nine wallets that hold half the vote.
- 80 / 80
- $23,000
- 58%
- 52%
- 56%
- 0
01 · The questionWho settles a bet when the two sides disagree
A Polymarket market is a yes/no question with a rulebook. When the event is over, anyone can propose the answer and post a bond. If nobody objects within a window, the market pays out. If someone does object, they post a bond of their own, and the question leaves Polymarket entirely. It goes to UMA, a second protocol on Ethereum. There, holders of the UMA token vote on the right answer. Votes are weighted by tokens. The majority wins, the winners earn rewards, and the losers are slashed.
Last week, Bloomberg reported that about half of that voting power sat with nine unnamed wallets, and that they almost always landed on the winning side. The Wall Street Journal had earlier found that in nearly one dispute in five, at least one voter also had money on the market being decided. On X, the line that stuck was "fewer people watch who decides the outcome." Polymarket's own answer was Protocol V2, unveiled on 5 October, which lets each market pick its oracle instead of UMA by default.
We wanted to know what that looks like right now, from the chain, at both ends. So we pulled every dispute from Polymarket's oracle contract on Polygon for the last month, every vote reveal from UMA's voting contract on Ethereum, and joined the two. The vote turned out to be the quiet half. The money moves at the first step, when someone posts a bond to object.
| What we read | Count |
|---|---|
| UMA voting rounds, 9 September to 7 October 2026 | 15, the first one partial |
| Vote reveals read from Ethereum | 218,384 |
| Questions put to a UMA vote and resolved | 489 |
| Of those matched to a dispute on Polymarket's oracle | 470 |
| Wallets that voted in a typical round | 440 to 470 |
| Disputes filed on Polymarket's oracle, 6 September to 8 October | 516 |
Polymarket runs on Polygon, and Bitquery indexes its markets, trades and payouts there. UMA runs on Ethereum, which we also index. The join between the two is a hash: each vote on Ethereum carries a stamp of the Polygon market it is about. Nineteen of the votes did not match. They reached UMA through other oracle contracts, seventeen on Polygon, one on Base and one on Ethereum. We count them in the voting figures and leave them out of the dispute tables.
02 · Who disputesSeventy-four wallets, one that never loses
Filing a dispute costs a bond, 500 USDC on Polymarket's oracle. Win and you get it back plus a cut of the proposer's bond. Lose and it is gone. That makes disputing a job, and a few dozen wallets do it. Every bond, answer and payout is visible in the settlement data.
Seventy-four wallets filed the 470 disputes we could match. The top five filed 42% of them. The most active, 0xc816f934fab8179997a05e4b71f405b9e2f69eb8, filed 80 and won all 80. Seventy-two of its wins were "too early" rulings: it watches for answers that land before the event is over, and disputes them. That is a fair public service. It is also steady income from other people's haste.
The bots it preys on are even more bunched up. Fifty-one wallets had an answer disputed; three of them account for 58% of the total. All three look like bots. The biggest, 0x4a1df1565a7704bbfa61e9cd543dd0265cc48bc0, has moved tokens about 549,000 times since March, had 126 proposals disputed, and was upheld on 48 of them. On X in late September, a digest of UMA's Discord named two other wallets as "repeatedly making incorrect proposals at off-hours to avoid disputers". One of them appears in our data with three disputed proposals, upheld on one.
| Proposers whose answers were disputed most | Disputed / upheld |
|---|---|
| 0x4a1d…8bc0 | 126 / 48 |
| 0x94f7…6faf | 81 / 4 |
| 0x89e4…a8b5 | 64 / 2 |
| 0x31e5…3eb4 | 21 / 2 |
| 0xd9aa…b5d3 | 20 / 0 |
03 · One night in Montevideo
Forty-six disputes, forty-two minutes, no wins
Late on 28 September, Montevideo Wanderers played Albion in Uruguay's top league. Polymarket had dozens of markets on the match: the winner, five spread lines, six totals, exact scores, second-half totals for each side. A proposer bot, the same 0x4a1df1565a7704bbfa61e9cd543dd0265cc48bc0, posted results for all of them.
Between 03:25 and 04:07 UTC the next morning, four wallets disputed 46 of those proposals. Each dispute posted a 500 USDC bond. Two days later UMA voted, and upheld the proposer on every one, with 94% or more of the weight each time.
We do not know why. The disputes were spread across four wallets in a pattern that looks scripted, and nothing in the match record suggests a real argument about the score. Whatever the cause, it reshaped the chart above. One of the four, 0xe0d478a2d0ee970fcf606649a17a2cc7544e8f48, filed 32 over the month and won eight. Twenty-two of its losses were this match. Another, 0x9a3fa403a6666eef75f92f181fcf13f9c051914a, filed 14 and won none. Take Montevideo out and the disputer side is a small set of pros who are almost never wrong. What were they so sure about? Mostly the clock.
04 · Too earlyHalf of all disputes are about the clock
UMA's vote has a fourth option besides yes, no and a split. It is called "too early", and it means the proposer jumped the gun: the match was not over, the day had not ended, the source had not published. The market goes back to Polymarket to be proposed again later.
More than half of all disputes in our window ended that way. Only a minority flipped the proposed answer to the opposite one.
| How 470 disputed markets resolved | Count |
|---|---|
| Too early, back to the market | 243 |
| No | 116 |
| Yes | 85 |
| 50/50 split | 26 |
| Disputer won (any of the above that differed from the proposal) | 375 |
| Of which an outright reversal, yes to no or no to yes | 132 |
| Proposer upheld | 95 |
The subject matter is dull too. More than four in five disputed markets were sports or esports props: set lines, half-time totals, map winners, exact scores. A handful were news markets, such as whether Iran charged fees in the Strait of Hormuz by the end of September, or whether a US-Iran truce held. The one crypto case that made the X timeline, the 5 October Bitcoin Up or Down market whose rules Polymarket changed and then undid, resolved at over 99% of the weight. Nobody on UMA was torn. Which raises the question Bloomberg asked: who is UMA, exactly?
05 · The nineNine wallets, every round
UMA runs a voting round every two days, and the disputes above are what fills it. A round covers whatever disputes arrived since the last one, from ten questions to over seventy. Round after round, the same picture.
The top wallet holds a sixth of all the weight on its own. The top three hold a third. The top nine hold just over half. Those shares barely move between rounds, because the same wallets show up every time with the same stake. Thirteen wallets rotate through the top nine over the month; four of them are in it in every single round.
| Voting weight, average of 15 rounds | Share |
|---|---|
| Top wallet | 17.0% |
| Top 3 wallets | 31.8% |
| Top 9 wallets | 56.3% |
| Median voter (the one in the middle when you line them all up) | about 1,037 UMA |
| Top wallet's stake | 3.26 million UMA |
| Wallets in a typical round | 440 to 470 |
None of the top wallets carries a label in any of the address lists we use. Three of the twenty-five biggest, one of them the second-largest, were funded through the same batch-send contract on Ethereum. That contract is Disperse, a public tool anyone can use, so on its own it proves little. Two got their first ETH in the same transfer, on 4 November 2024, from the same signer. That signer, 0x774bff0ccd2f7546b67d4071def9ba154ba549b1, is unlabeled too. It is the only visible thread between any of the big voters, and it links two of them.
06 · Do they decide?A block that agrees with everyone
The nine biggest wallets, ranked by their peak stake in the month, do vote as a block. On 483 of the 489 votes, every one of them that voted cast the same answer. And they win: each of the nine was on the winning side in at least 99% of its votes.
But so was almost everyone. The median voter was on the winning side 89% of the time. Only two votes in the whole month finished with the winner under 75% of the weight. Both were tennis over/unders.
We then re-ran every vote with the top nine removed. The result changed on none of them. Zero of 489.
| The top 9, across 489 resolved votes | Result |
|---|---|
| Votes where all of them cast the same answer | 483 |
| Lowest win rate among the nine | 99% |
| Median win rate, all voters with 20+ votes | 89% |
| Votes where the winner held under 75% of weight | 2 |
| Votes where the winner held under 90% of weight | 5 |
| Results that change if the top 9 are removed | 0 |
So Bloomberg's number holds, and so does the worry behind it: nine wallets could decide any contested market they wanted to. In this month they did not need to. The calls were easy, the bonds did the sorting, and the only people who lost real money were the ones who filed 46 disputes on a football match.
07 · What changesThe vote is about to matter less
Polymarket's Protocol V2, due to switch on for new markets from 2 November if the date holds, routes each market through an aggregator that can point at UMA, Chainlink, or another source. Sports, crypto and weather are expected to leave UMA first. Those are exactly the categories that fill UMA's queue today. If they go, the vote will be left with the news markets, the ones where a result can turn on how one word in the rules is read.
That is where concentration would bite. In this month the nine never had to use their weight, because the calls were easy. On a hard one, they will.
Trace the next dispute
Every figure here comes from two Bitquery datasets: Polymarket's trades, disputes and payouts on Polygon, and UMA's vote reveals on Ethereum. Both are open through the same API, so you can pull the next round's votes, check a disputer's record, or follow a bond.
FAQ
Who decides a disputed Polymarket market?
UMA token holders, by weighted vote on Ethereum. Polymarket's own oracle contract on Polygon sends the question to UMA when someone posts a bond against the proposed answer.
Do nine wallets really control the vote?
They hold 56% of the weight, so yes, together they could decide any vote. In the 30 days we checked they never needed to: removing them changes none of the 489 results.
What does "too early" mean in a UMA vote?
That the proposer submitted an answer before the question could be settled. The market returns to Polymarket for a fresh proposal. It was the outcome of 52% of disputes in our window.
How much does it cost to dispute a Polymarket market?
A 500 USDC bond on Polymarket's oracle. The disputer gets it back with a share of the proposer's bond if UMA agrees, and loses it if not.
Will Protocol V2 change this?
New markets can choose their oracle from 2 November 2026, if the date holds. Sports, crypto and weather markets are expected to move away from UMA first, which would remove most of today's dispute volume from the vote.
Vote weights are the token amounts each wallet revealed in a round. A wallet is not a person, and several of the top nine may belong to one holder, or to a staking service voting on behalf of many.
The win rate of a voter is whether its revealed answer matched the resolved answer. It says nothing about whether the voter was right about the world.
Disputes were matched to votes by hashing the Polygon question text with the requester address, the way UMA's contracts do. 470 of 489 votes matched; the other 19 came through different oracle contracts and may not be Polymarket markets.
A disputer 'won' if the resolved answer differed from the proposed one, including a 'too early' ruling. The bond in all 46 Montevideo dispute transactions was read from the chain: 500 USDC each. The total assumes each lost dispute forfeited the full bond; we did not trace the payouts at settlement.
Two of the top voters were funded in one transfer by the same signer. A third used the same public batch-sending tool, which is not a link on its own. Shared funding is not proof of common ownership.
The proposer and disputer wallets are described by what they did on-chain. We do not know who runs them.
Round 10367 was still in its reveal phase when we measured and is excluded. The first round in the window is partial.
In 15 of the 470 disputes the proposer and the disputer were the same wallet. They are counted like any other dispute.
This article is provided for informational and educational purposes only and reflects analysis of publicly available on-chain data as of the dates indicated. It does not constitute legal, financial, compliance, or investment advice, and nothing in it is a recommendation to buy, sell, or hold any token or asset.
Wallet addresses named here are identified by their on-chain behaviour only. No claim is made about the identity, intent or legal status of any person or entity behind them, and the pattern called a block or a bot is an inference from timing and funding, not a finding of coordination.
References to Polymarket and UMA describe what their public contracts recorded. Nothing here asserts that either protocol, or any voter, proposer or disputer, acted improperly. Prediction markets are restricted or unlawful in some jurisdictions.
Nothing herein should be relied upon as a definitive determination of fact. Readers should conduct their own independent verification before taking any action. The authors and publisher accept no liability for any loss or damage arising from reliance on this material. All trademarks and company names are the property of their respective owners.
Reported by Divyasshree for Bitquery Research, with AI tools; every figure was checked against the raw data.