Every prediction market has to answer one question after every trade closes: what actually happened?
Was BTC above $100K on December 31? Did Trump win Pennsylvania? Did Thailand strike Cambodia by August 15? A yes or no. A one or a zero. A payout, or a loss.
Getting that answer right — and doing it in a way that doesn't get gamed — is the hardest problem in the industry. It's called the oracle problem, and there are three completely different approaches to solving it in 2026. Here's how each works, when to trust which, and why oracle design is the single most under-discussed feature of the platforms you're trading on.
The oracle problem, stated simply
A prediction market is a smart contract. Smart contracts execute deterministically based on inputs. But smart contracts can't see the real world. They don't know what BTC's price is, who won an election, whether a war has begun. That information has to be fed in from outside — by something called an oracle.
The oracle problem is: how do you feed real-world information into a smart contract in a way that's accurate, timely, and can't be manipulated?
Every prediction market platform solves this differently. The three main approaches are UMA (Polymarket), Chainlink (World, and increasingly others), and expert panels (Kalshi). Each has fundamentally different strengths and weaknesses.
Approach 1: UMA — Optimistic Oracle voting
UMA (Universal Market Access) is the oracle Polymarket uses. It's a decentralized voting system for resolving disputed outcomes.
Here's how it works. When a Polymarket market is ready to resolve, anyone can propose a resolution ("YES, BTC closed above $100K"). This proposal sits in a challenge window, usually 2 hours. If no one disputes, the resolution stands and the market pays out.
If someone disputes, the question goes to a vote of UMA token holders. They see the question and the evidence, and they vote on the correct outcome. The winning side of the vote gets a small reward. The losing side loses a small amount of UMA tokens. Economic incentives should push honest voting.
Advantages: extremely flexible. UMA can resolve any question — subjective, ambiguous, contested. It can handle "Did event X count as a war?" questions that machines can't. Political and geopolitical markets thrive on this flexibility.
Disadvantages: slow. Voting takes days for disputed markets. Also potentially manipulable — if a whale accumulates enough UMA tokens, they can influence outcomes. This exact concern has been raised in the 2024 US election resolution disputes.
Approach 2: Chainlink — Automated data feeds
Chainlink is a decentralized oracle network that aggregates data from multiple sources and delivers it to smart contracts. It's the oracle that "World" uses, and it's spreading to other prediction market platforms.
How it works. Each market specifies which data source it reads from — say, Coinbase's BTC price at midnight UTC on December 31, or the Associated Press's declared winner of a state election. Chainlink's decentralized network of nodes reads that source, verifies consensus among nodes, and pushes the result to the smart contract. The contract executes automatically.
Advantages: fast. Resolution happens within minutes of the underlying event. Also tamper-resistant — you'd have to attack multiple independent Chainlink nodes simultaneously to manipulate an outcome. And deterministic — there's no voting or subjectivity.
Disadvantages: rigid. Chainlink can only resolve questions where a machine-readable data source exists. "Did BTC close above $100K?" — easy. "Did Powell say 'transitory' in his press conference?" — extremely hard. "Did the situation in Iran count as a war?" — essentially impossible.
Approach 3: Kalshi — Internal expert resolution
Kalshi is a centralized, CFTC-regulated exchange. Its approach to the oracle problem is dramatically different from the on-chain platforms: they use internal experts.
How it works. Each Kalshi market has a specified resolution methodology written in the contract listing. When the event occurs, Kalshi's internal operations team reviews the evidence, applies the methodology, and posts the outcome. Users can dispute through customer service, but final decisions rest with Kalshi.
Advantages: extremely reliable. No voting attacks. No smart contract bugs. Full accountability — Kalshi as a company is legally responsible for correct resolution. Regulatory-friendly, which is why Kalshi got CFTC approval.
Disadvantages: centralized. You have to trust Kalshi. There's no code that guarantees correct resolution. Also less transparent — the internal review process is not fully public. This has caused controversies, including a January 2026 incident where Kalshi initially refused to pay out on certain NFL bets before reversing under public pressure.
The trade-offs in one table
UMA: flexible, decentralized, but slow and manipulable. Best for: political, geopolitical, subjective questions.
Chainlink: fast, tamper-resistant, but rigid. Best for: price-based markets, sports outcomes with clear scoring, macro data releases.
Expert (Kalshi): reliable, accountable, but centralized and requires trust. Best for: regulated markets that need to survive legal scrutiny.
No approach dominates. Each is optimal for different types of markets. This is why the industry is fragmenting by oracle type — because different questions need different resolution machinery.
The disputes that have actually happened
Three case studies from 2024-2026 that illustrate what oracle design gets right and wrong.
The 2024 US election. Polymarket used UMA to resolve state-level electoral markets. Several states had contested outcomes that required voting. Most resolved cleanly. A few — particularly Pennsylvania's initial ambiguity — took days to settle. Trader complaints followed. UMA held up under stress, but not smoothly.
The 2025 Iranian leader death market. Kalshi's contract on whether Iran's Supreme Leader would be "out" by certain dates was frozen after Israeli airstrikes killed him. Kalshi refused to resolve, citing internal policy against "transactions directly tied to death." $54M in positions were locked. This was Kalshi's expert-panel discretion overriding the contract's stated resolution mechanism. Users lost trust.
The 2026 NFL Kalshi refund incident. Kalshi initially paid winners only their original stake, not the full winnings, on certain player prop markets. Only after public backlash did they reverse and pay out fully. Again — centralized discretion, weakly bounded by contract terms.
Each incident is a lesson in oracle failure modes. UMA is slow. Chainlink is inflexible. Expert panels are unaccountable. There is no perfect oracle.
What good oracle design looks like
The next generation of prediction market platforms is trying to combine the best of all three approaches.
Hybrid oracles: use Chainlink for machine-readable resolutions (prices, sports scores) and UMA-style voting only for edge cases. This preserves speed while retaining flexibility. Several new platforms are experimenting with this approach.
Contract-specified resolution sources: every market specifies exactly which data source resolves it, published upfront. No expert discretion, no post-hoc reinterpretation. If the source says X, the market says X.
Backup dispute mechanisms: even automated resolutions can have edge cases (data source outage, ambiguity, wrong number reported). Good oracle design includes explicit fallback procedures that are transparent to users.
Time-bounded challenges: resolution disputes have to be raised within X days of the event, not months later. This prevents post-hoc gaming.
What Thai users specifically should know
For Thai users considering prediction markets, oracle design matters more than usual for two reasons.
First, Thai-specific markets are harder to oracle. There's no Chainlink price feed for the SET Index (yet). There's no automated news source that would resolve "Did Thailand strike Cambodia by August 15?" in a machine-readable format. Any Thai-focused market has to either rely on UMA-style voting (with its manipulation risks) or expert panels (with their accountability concerns).
Second, Thai users' recourse if oracle failure occurs is limited. Polymarket has no Thai regulatory presence. Kalshi is inaccessible. World is on Solana with limited support infrastructure. If a resolution goes wrong, there's no Thai regulator to appeal to.
A Thai-domiciled prediction market with local regulatory oversight solves this. Oracle disputes could be adjudicated through Thai courts. Data sources could be from Thai financial institutions. Recourse would exist.
The three oracles you need to understand
If you take one thing from this piece: before you trade on any prediction market, know which oracle resolves your contract.
UMA-resolved contracts: expect delays if the question is contested. Consider that resolution ultimately depends on token-holder voting. Be aware that manipulation is theoretically possible.
Chainlink-resolved contracts: expect fast, clean resolution. Verify the data source is reasonable and won't have outages. Understand that ambiguous or subjective outcomes may not resolve at all.
Expert-resolved contracts: read the contract terms carefully. Trust the platform, or don't trade there. Know that resolution is discretionary within stated methodology.
The oracle isn't a technical footnote. It's the mechanism that determines whether you get paid. Understand it before you trade.
What Juno is building for this
Juno's oracle strategy is hybrid by design. Machine-readable markets (price, macroeconomic data) resolve automatically via Chainlink-style feeds. Subjective markets use a curated expert panel with published methodology, transparent decision logs, and time-bounded dispute periods.
For Thai-specific data sources, Juno integrates directly with Thai financial data providers and news sources — not stateless global feeds. This means "Will BoT cut rates?" can be resolved cleanly from official BoT publications, not through voting or discretion.
The design principle: fast when the answer is unambiguous, accountable when it isn't. That's what the next generation of prediction market oracles needs to look like.
The one-word lesson
Oracles are the invisible layer that makes prediction markets work. Get the oracle right and the market is trustworthy. Get it wrong and the whole platform collapses regardless of how good the interface looks.
Polymarket has UMA. World has Chainlink. Kalshi has experts. Each has strengths, each has weaknesses, and understanding which is powering your contract is the difference between a market you can trust and a market you're just hoping resolves correctly.
The next time you click "Buy YES," check the oracle. It's the most important line item on the contract that nobody reads.