Original | Odaily Planet Daily (@OdailyChina)
Author | Azuma (@azuma_eth)
Polymarket has once again found itself embroiled in controversy over fairness.
The incident originated from the prediction market "Will the U.S. invade Venezuela by...?". On January 4th, Polymarket intervened to add a clarification stating that "the previous U.S. operation to capture Venezuelan President Maduro does not meet the definition of an invasion." This announcement caused a sharp drop in the price of YES shares (betting that the U.S. would invade Venezuela before January 31st) in this prediction pool, directly impacting the actual interests of many users.
- Odaily Note: The chart shows the price trend of YES shares betting on January 31st. The turning point is the moment when the official Polymarket account intervened to add the clarification.
This is not the first time Polymarket has faced similar controversy. Last year, in our articles "Polymarket Suffers Oracle Manipulation Attack, Can Whales Use Voting Power to 'Reverse Black and White'?" and "Polymarket Faces Another Truth Dispute: Zelenskyy's Outfit Will Determine the Fate of $140 Million," we discussed similar cases and briefly analyzed Polymarket's result adjudication logic.
During the discussion of this recent event, we noticed that although many readers are aware that Polymarket relies on the oracle protocol UMA for adjudication, they are not clear on how this process operates. Therefore, Odaily is publishing another article to parse its adjudication mechanism and attempt to explore the ambiguous areas that could potentially lead to disputed outcomes.
Predefined Rules and Supplementary Explanations
First, any prediction pool on Polymarket is created with a set of predefined rules. These rules clearly state the conditions for determining the outcome, the validity period, and anticipate in advance how to adjudicate under various unexpected circumstances.
Take the prediction pool "Will the U.S. invade Venezuela by...?" as an example. As shown in the image above, the text under "Rules" is the predefined rule for this pool. The judgment condition and validity period are — if the United States launches a military offensive aimed at controlling any part of Venezuelan territory between November 3, 2025, and January 31, 2026 (11:59 PM EST), the outcome will be adjudicated as YES; otherwise, it will be NO.
However, even with various contingencies anticipated in advance, sometimes events unfold in unexpected ways. For instance, in this case, no one could have predicted that a country's president could be so suddenly captured by another force. Therefore, in very rare cases, Polymarket will personally intervene to provide supplementary explanations for unexpected situations not foreseen when the market was created, further clarifying the rules — the decision to clarify is not solely made unilaterally by Polymarket; users with doubts can actively request clarification in the #market-review channel of the Polymarket Discord.
Observant readers may have noticed that below the "Rules" in the image above, there is a fainter section labeled "Additional context," with a more recent update date (the predefined rule was posted on December 18th last year, while this content was added on January 4th). This is precisely the content Polymarket added as a supplementary explanation this time. The specific content is — "This market concerns US military action aimed at establishing control. President Trump, referring to ongoing negotiations with the Venezuelan government, stated he would 'manage' Venezuela, but this statement alone is not sufficient to characterize a 'capture and extraction' mission aimed at apprehending Maduro as an invasion."
Simply put, Polymarket does not believe the U.S. capture of Maduro should be defined as an invasion of Venezuela, and therefore does not support adjudicating the outcome as YES based on this event.
Let's set aside whether Polymarket's supplementary explanation is reasonable for now. What's more important to note here is that the validity period of this prediction pool (January 31st) has not yet ended, meaning it has not entered the final adjudication process. The reason for emphasizing this is twofold: first, to remind everyone that all current disputes essentially stem from rule ambiguities, unrelated to the adjudication环节 itself; second, to illustrate that this dispute is not yet settled, and users' current losses are actually paper losses—everything depends on the final adjudication outcome.
So how is the final adjudication process executed?
Adjudication Process: Outcomes Are Proposed by People
For any prediction pool on Polymarket, during the final adjudication process, someone must propose an outcome. Taking the previous prediction pool as an example, the window for proposing an outcome is right under "Rules" at "Propose resolution."
Of course, not everyone can casually propose an outcome胡乱. UMA and Polymarket have designed two main restrictions here: economic incentives and a whitelist requirement.
The economic incentive means that proposing an outcome requires submitting a amount of USDC as collateral (usually 750 USDC, higher for some pools). After submission, there is a challenge window (usually 2 hours). If no one challenges during this period, the outcome is deemed valid and will be used as the basis for the final adjudication of the prediction pool, with no further changes. The proposer then receives a certain bonus (usually 5 USDC). Conversely, if challenged, it enters a dispute phase, and the proposer risks losing their collateral (detailed below). Simply put, if someone proposes a random outcome just to cause trouble, the risk far outweighs the reward.
- Odaily Note: Clicking on "Propose resolution" in the prediction pool shows the collateral requirement and bonus amount for proposing an outcome.
The whitelist restriction means that Polymarket initially allowed anyone to propose resolutions, but later, to improve adjudication efficiency, introduced a whitelist maintained jointly with Risk Labs in August last year. Subsequently, only whitelisted addresses are allowed to propose outcomes. There are three ways to get on the whitelist: first, join the Risk Labs team; second, join the Polymarket team; third, have submitted over 20 proposals with an accuracy rate exceeding 95% in the past three months — all addresses can be queried through this contract. Initially, there were only 40 addresses, but the number has now expanded significantly.
Dispute Phase: Economic Interest Game
As mentioned in the previous part, if a proposed outcome receives no异议 during the challenge window, it is deemed valid. This is the final outcome for the vast majority of prediction pools. However, in very rare cases, if an objection is raised, how is the adjudication made?
First, it must be said that, like proposing an outcome, objections cannot be raised frivolously — the objector must pay an equal amount of USDC as collateral (usually still 750 USDC) to confront the proposer, meaning both parties must put an equal stake on the table. However, unlike the proposer, the objector does not need to provide a complete alternative outcome; they only need to point out a specific error in the proposer's outcome.
Once an objection is confirmed, the UMA community will debate it. This phase typically lasts 24-48 hours (voting occurs the next day, allowing at least 24 hours for discussion each time). Anyone wishing to provide evidence for the discussion can do so in the #evidence-rationale and #voting-discussion channels of the UMA Discord server.
After the debate, UMA token holders will vote on the matter (this process takes approximately another 48 hours), and one of the following four outcomes may occur:
- Proposer Wins: The proposer retrieves their collateral plus half of the objector's collateral as a bounty. The objector loses their collateral.
- Objector Wins: The objector retrieves their collateral plus half of the proposer's collateral as a bounty. The proposer loses their collateral.
- Too Early: This outcome applies to proposals where the relevant event has not yet concluded, such as an ongoing sports game result. The objector receives a refund plus half of the proposer's collateral as a bounty. The proposer loses their collateral.
- Draw (50:50): The rarest scenario. In this case, the objector retrieves their collateral and receives half of the proposer's collateral as a bounty. The proposer loses their collateral.
Two points need attention in the above voting.
First, among the four potential outcomes, the objector profits in three scenarios, while the proposer profits in only one — this is intentional design by UMA, aiming to increase the accuracy of proposed outcomes through an imbalance of risk and reward between the parties. Since the objector only needs to point out one flaw to win, the proposer must provide a result that is as accurate and standard-compliant as possible.
The second point is that UMA's governance voting power holds absolute authority over the final outcome. In other words, the billion-dollar prediction market spectacle built by Polymarket is ultimately supported at its foundation by a protocol with a Fully Diluted Valuation (FDV) of only $100 million.
Exploring the Ambiguous Zones
Combining the above analysis of Polymarket's adjudication process with a review of historical real dispute cases, it is not difficult to find that there are certain ambiguous zones that can cause controversy, both during the rule-setting and supplementary explanation phase of the prediction pool's operation and in the final adjudication process.
First, in the rule-setting and supplementary explanation phase, the essence lies both in the fact that written rules sometimes cannot cover real-world variables, and in the fact that the same textual description can often be interpreted in different ways. For example, last year's "Did Zelenskyy wear a suit?" incident: firstly, the rules did not specify whether a "military-style suit counts as a suit"; although Polymarket explained in a supplementary clarification that "reliable reports have not confirmed whether Zelenskyy wore a suit," it did not define what constitutes a "reliable report." Ambiguities like these are destined to cause controversy.
If Polymarket itself, as a platform, could remain neutral, it might not anger the public every time, but the situation is hardly ideal. Polymarket's operating entity is based in the United States, which means the regulatory environment and political context it faces make it difficult to remain completely neutral on all issues involving geopolitics. For instance, in the current case of "Will the U.S. invade Venezuela?", when it comes to U.S. military and diplomatic actions themselves, rule interpretations tend to lean towards more conservative, "non-militarized" descriptions. This is not incomprehensible, but the ultimate losers are the users.
As for the adjudication process, the source of ambiguity points directly to the potential for fraud in UMA voting. Although UMA has designed a reward and punishment game mechanism to constrain proposal behavior and improve outcome accuracy, this game mechanism only constrains the internal economic interests within its system. When external profit opportunities exist, the potential for malicious activity theoretically remains. This is not baseless suspicion; in last year's "Ukrainian Rare Earth" incident, a UMA whale manipulated voting power to forcibly reverse the outcome, resulting in bets worth $7 million being adjudicated incorrectly.
The existence of these ambiguities is the root cause of frequent质疑 about Polymarket's fairness and is also a structural problem that prediction markets need to solve. In fact, any prediction market involving complex real-world events will inevitably face the following triple dilemma — First, real-world events themselves often cannot be clearly binarized; geopolitics, military actions, and diplomatic games are inherently full of gray areas. Second, rules must be expressed in language, but language naturally allows for interpretive space. Third, once human or governance participation is introduced into the adjudication mechanism, interest博弈 becomes inevitable.
From the user's perspective, perhaps you need to realize early on — in prediction markets, what you are betting on is not "what will happen in the world," but "how the rules will ultimately be interpreted."
