How do prediction market payouts work?

Most event contracts are structured to pay a fixed amount, often $1, if the defined outcome happens by expiration and $0 otherwise. If a trader bought YES at $0.70 and YES resolves, the share redeems at $1 and the gross gain is $0.30 per share. Losing shares expire worthless at $0.

How does Polymarket work compared with other prediction markets?

Polymarket is a crypto-native venue where users trade outcome positions and then rely on an oracle process to resolve the result. CoinLaw reports Polymarket shifted from an AMM to a CLOB in late 2022 to address liquidity issues in binary markets. CoinLaw also describes Polymarket using UMA’s optimistic oracle for resolution.

Are prediction markets more accurate than polls?

Prediction markets are often framed as information-aggregation tools because participants risk money and prices update in real time. Polls measure stated preferences at a point in time, while markets reflect tradable prices that can move quickly with new information. The key caveat is that market prices can be distorted by spreads, fees, and thin liquidity, so the signal is not always a pure forecast.

What does it mean when YES and NO do not add up to $1?

It usually means microstructure is dominating the neat theory. CoinLaw notes liquidity, supply and demand, and transaction costs can push YES and NO away from a perfect $1 sum. The deviation can look like arbitrage, but it is only real if both legs can be filled and the edge survives fees.

How do prediction markets work: Contracts, order books, and

Q: How are prediction market prices set?

Prices are set by supply and demand as traders post bids and asks and trade against each other. The last traded price is often read as an implied probability, but spreads, fees, and thin liquidity can distort that signal. YES and NO prices may not sum to exactly $1 because of transaction costs and lopsided order flow.

Prediction markets work by listing an event contract with defined outcomes, letting traders buy and sell positions whose prices move with supply and demand, and then settling to $1 or $0 once the outcome is resolved. The useful shortcut is that the trading price can be read as an implied probability, but the tradable reality is shaped by spreads, fees, liquidity, and the oracle process that decides how prediction markets resolve.

Key Takeaways

Prediction markets trade event contract payoffs, often as a binary contract that settles to $1 for the winning outcome and $0 for the loser.
A contract price can be read as an implied probability, but microstructure means the screen price is not a clean forecast.
Most crypto prediction markets gravitated to CLOBs because AMM liquidity providers can be forced into losses as one side goes to zero at resolution.
In decentralized venues, settlement is the product: smart contracts pay out only after an oracle or reporting system adjudicates the outcome.

Prediction markets as tradable probabilities

The contract that sits under most prediction markets is a simple payoff: $1 if a specific outcome happens by a deadline, $0 if it does not. That payoff is packaged as an event contract with clearly defined terms, then traded before expiration. The platform is not supposed to be “the house” taking the other side. The platform is a venue that lists the contract, matches buyers and sellers, and charges fees.

Once the payoff is fixed at $1, the price becomes a probability-like number. If YES trades at $0.70 on a $1 payout, the market is implicitly valuing the outcome at about a 70% chance. That heuristic is why prediction markets get compared to forecasting tools. The price is a live number that updates when new information hits and someone is willing to pay up or hit bids.

The important nuance is that the “price equals probability” shortcut is an interpretation layered on top of a trading instrument. The instrument is still a claim on a $1 settlement, and P&L is still entry price versus exit price, plus fees. That framing matters because it forces the right question: is the displayed probability a clean forecast, or is it a liquidity-weighted opinion that can be pushed around by thin books and wide spreads.

Crypto rails mostly change who can access the venue and how the contract is collateralized and settled. Many on-chain prediction markets use stablecoins as collateral so the wager is about the event, not about the collateral asset’s volatility. The broader prediction markets category is still the same machine: a defined payoff, a trading venue, and a settlement process.

How pricing and trading actually happen

Prices move because traders post bids and asks for the YES and NO sides and then trade against each other. The venue’s matching rules decide whose order fills first, and the last traded price becomes the number everyone quotes. That is the core of prediction market mechanics: the price is not “calculated” by a model, it is discovered by orders interacting.

A clean mental model is to treat each share like a $1 digital note that might be worth $1 later or might be worth $0. If a trader buys YES at $0.70 and the market later resolves YES, the position redeems at $1 and the gross gain is $0.30 per share. If the trader is wrong, the share goes to $0. That is also why defined risk can still feel brutal. The maximum loss is the premium paid, but the realized loss can be driven by bad exits when liquidity disappears.

“How are prediction market prices set” becomes tricky once microstructure shows up. CoinLaw flags a key screen-level tell: YES and NO prices may not sum to exactly $1 because liquidity, supply and demand, and transaction costs can create deviations. That means the displayed “probability” can be distorted by:

1. Spreads: a market can show $0.70 last trade while the best bid is $0.66 and the best ask is $0.74. 2. Fees: a small edge can vanish once the venue takes a cut on entry and exit. 3. Thin books: a single order can move the midpoint, so the price becomes a function of who is willing to show size.

Those deviations also create the classic temptation: “arb” when YES + NO is not $1. The sanity check is whether both legs can be filled at the displayed prices and whether the edge survives fees. Paper arbitrage is common in illiquid contracts because the screen shows theoretical relationships that cannot be traded in size.

Market design: CLOBs and AMMs

The matching engine decides whether the contract trades like an order-book instrument or like a pool swap. Two designs dominate: the central limit order book (CLOB) and the automated market maker (AMM). A CLOB is the familiar stack of bids and asks where traders choose their prices and get priority by price and then time. An AMM is a smart-contract pool that quotes a price algorithmically based on the pool’s inventory.

Binary settlement changes the economics in a way many DeFi-native traders underestimate. CoinLaw’s claim is blunt: most crypto prediction markets use a CLOB rather than AMMs because binary outcome tokens settle to $1 or $0, so one side of an AMM pool goes to zero at resolution. That is why LP losses can feel “guaranteed” unless fees are enormous. The pool is effectively warehousing a position that gets dragged toward the winning side as information arrives, and then one inventory line item becomes worthless at settlement.

This is also where the “short gamma” intuition comes from. An AMM that is always willing to quote both sides is mechanically selling convexity into a terminal $0/$1 jump. As the market trends toward the eventual winner, the pool’s inventory shifts in a way that leaves it holding more of the losing token into resolution. The LP is not just earning fees. The LP is taking a path-dependent risk that is structurally hostile in a binary payoff.

“How does Polymarket work” is a useful concrete example because CoinLaw reports Polymarket moved from an AMM model to a CLOB in late 2022 to address liquidity issues with binary outcome tokens. That shift is a tell about what scales: once traders care about spreads and exits, order books tend to be the natural fit for event risk.

ChainUp presents AMMs as a key component for continuous liquidity in crypto prediction markets, and that is directionally true for long-tail markets that would otherwise be dead. The trade-off is that “always-on liquidity” is not free. Someone is paying for it, and in binary settlement that someone is often the LP.

Resolution and payout after the event

Everything before expiry is just trading a claim. The contract only becomes money when the venue decides the outcome and settles. How prediction markets resolve is the part that looks boring until it breaks, because it determines whether the $1 payoff is real.

The payout math is simple. A winning YES share in a $1 binary contract redeems for $1 and the losing side redeems for $0. The hard part is deciding which side is “winning” in a way that matches the contract’s wording. Centralized venues handle this with a rulebook and an operator that declares the result based on predefined sources. Decentralized venues need an oracle layer because smart contracts cannot observe real-world facts on their own.

CoinLaw describes Polymarket using UMA’s optimistic oracle, which runs a request–propose–dispute process. The fastest path is about 2 hours when nobody disputes the proposed outcome. If a dispute happens, the process escalates, and CoinLaw notes it can end in UMA token-holder voting via the Data Verification Mechanism.

That structure creates a very specific settlement profile:

1. A result is proposed on-chain. 2. If nobody challenges it inside the window, the market finalizes quickly. 3. If challenged, the market’s “truth” becomes a governance process with incentives and voting power.

ChainUp also emphasizes that decentralized markets require oracles to bring outcomes on-chain so smart contracts can pay winners. That is the correct mental model: the smart contract is the cashier, not the judge. The judge is the oracle process.

Decentralization choices and key risks

Decentralization is often sold as “no trust.” The actual shift is that trust moves from an operator to an incentive system and its governance. The Augur whitepaper motivates decentralization by pointing at the fragility of centralized operators and the problem of concentrating power, including the line, “Power corrupts. Absolute power corrupts absolutely.” The design goal is censorship resistance and a settlement process that does not rely on a single judge.

Augur’s approach is to make reporting itself a token-incentivized activity. The whitepaper introduces a tradeable Reputation token with fixed total supply. Holders gain or lose Reputation based on whether they vote with consensus, and they are obligated to vote at periodic check-ins, default every 8 weeks. That is a mechanism for keeping reporters engaged and making dishonesty expensive.

The failure mode is governance concentration. CoinLaw points to a concrete example: a roughly $7 million Polymarket contract that falsely settled in March 2025 via concentrated UMA voting power. That is the cleanest illustration of the thesis risk. Being “right” about the real-world outcome is not sufficient if the adjudication layer can be pushed to a different answer.

Liquidity is the other risk that shows up on a trader’s screen. Defined-risk instruments can still behave like illiquid alts. When the book thins, the probability signal becomes less informative and the exit price becomes the whole game. This is also where the “YES + NO not equal $1” phenomenon stops being a curiosity and starts being a warning that the market is not deep enough to enforce tight relationships.

Near expiry, the market’s center of gravity shifts from forecasting to settlement. The most important document becomes the contract’s exact event definition and timing. Ambiguous wording is where disputes are born, and disputes are where time-to-cash stretches out.

A full lifecycle example in Augur

Augur’s whitepaper breaks the lifecycle into two phases: before and after the event. The walkthrough starts with a user creating an event, then creating a market that contains that event, then letting shares trade during the forecasting phase, and finally resolving after the event occurs.

Event creation is not just typing a question. The creator specifies fields that function like a term sheet: description, type (binary, scalar, or categorical), the valid range of answers, a fee to create the event, and the maturation and expiration timing. Time is represented by block intervals, and the user can input either a block number or an approximate end time. The event is also tagged with a topic category, and the creator’s address is part of the event data.

Market creation then packages one or more events into a tradable venue and seeds liquidity. The whitepaper describes up-front funding provided by the market’s creator, which covers the creator’s maximum possible loss in that market. It also defines a loss limit parameter that controls liquidity versus the creator’s potential loss, with maximum loss scaling with the number of outcomes. The point is not the exact formula. The point is that liquidity is paid for, explicitly, by someone taking a bounded loss profile.

From there, trading happens during the maturation window. Shares are created for the events in the market, and participants buy and sell those shares as information changes. After expiration, the system transitions into the reporting and resolution phase, where the network “checks in” with reality and Reputation holders vote.

That end-to-end design makes the three-part plumbing diagram visible: the contract encodes a $1 payoff, the trading layer discovers a price that traders read as implied probability, and the settlement layer decides what reality means for the contract. The broader prediction markets idea only works when all three layers are credible at the same time.

The Take

I’ve watched traders treat a $0.70 YES like it’s a clean 70% forecast, then get clipped by the boring stuff: a wide book, fees that eat the edge, and a settlement definition that was looser than it looked. The trade is easy to understand because it’s a $1 payoff. The hard part is remembering that the “probability” on the screen is a tradable number, not a truth oracle.

I’ve also seen the settlement layer become the whole story. CoinLaw’s March 2025 example of a roughly $7 million Polymarket contract falsely settling via concentrated UMA voting power is the kind of failure mode that changes how these products should be read. Being wrong is one risk. Being right and still losing because the adjudication layer says otherwise is the one that actually breaks people.

How do prediction markets work: Contracts, order books, and settlement