On this page
- What does a prediction market trade?
- What does “decentralized” describe?
- Why is a price of 0.60 often read as 60%?
- Separate the payout from the profit
- How does a market probability differ from a poll?
- Four checks before you use a percentage
- How can forecast quality be evaluated?
- Common questions
- Read the question before the number
A prediction market lets people buy and sell contracts whose payouts depend on the outcome of an event. Its price can tell you something about expectations, but you need the contract rules to interpret it. A price described as “60%” is neither a promise that the event will happen nor a poll showing that 60% of people support it.
What does a prediction market trade?
Consider a contract asking whether an economic indicator will exceed a specified threshold. A simple binary contract might pay one unit for each YES contract if the answer is yes, and zero if the answer is no. People with different views buy and sell those contracts, creating a price you can observe. Binary fixed-payout contracts are a common example; other designs also exist. Research on prediction-market designs
“Prediction market” describes a mechanism. Polymarket is one platform using that mechanism. Some markets use real money and others use points; cash redemption, settlement assets and participation restrictions depend on the rules. The price of a cryptocurrency associated with a prediction-market project is also different from the price of a contract on a particular event.
The question matters as much as the number. A contract about a government announcing a policy and one about that policy taking effect can resolve differently, even if their headlines look similar. Compare their conditions before comparing percentages.
What does “decentralized” describe?
A prediction market can use blockchain for some functions while relying on an operator for others. Ask who matches orders, where trades settle and how the event’s outcome is decided.
For example, Polymarket Global’s order-lifecycle documentation describes a hybrid system:
- Off-chain matching: Users sign orders, and an operator checks them and matches compatible orders in the order book.
- On-chain settlement: Matched trades are submitted to smart contracts that verify the orders and settle the trade on Polygon.
Deciding which outcome won is a separate function from matching or settling a trade. The platform’s resolution documentation describes UMA’s oracle process for prediction markets and Chainlink time-weighted average price (TWAP) data for up-down markets. Read the individual market’s rules rather than assuming every contract uses the same process.
On-chain settlement therefore does not mean the service has no operator or that outcome data and dispute procedures no longer matter. Nor does the architecture label alone establish custody arrangements, insurance or your eligibility to trade. The example describes the platform’s published design; it is not an independent security assessment.
Why is a price of 0.60 often read as 60%?
For a simplified contract paying one unit on YES and zero on NO, a price of 0.60 is 60% of the maximum payout. It is therefore often described as an implied probability of about 60%. That is an interpretation of a contract price, rather than a measurement of an objective, fixed probability.
Research by Justin Wolfers and Eric Zitzewitz examines when prices equal or approximate traders’ average beliefs. That relationship depends on assumptions, including risk preferences and the distribution of beliefs and wealth. Prices can be informative while still being biased or differing from those beliefs. How prices relate to probabilities
When you see a percentage, check which contract and time it describes. A last trade is different from the current bid or ask. A displayed price also does not guarantee that the quantity you want can trade at that price.
Separate the payout from the profit
This is a hypothetical example using one common unit of account, not a live quote from any platform. Suppose all 100 YES contracts fill at 0.60 each, for a total cost of 60. You hold them to a normal YES/NO resolution: each purchased YES contract pays one unit on YES and zero on NO.
| Item | Resolves YES | Resolves NO |
|---|---|---|
| Purchase cost | 100 × 0.60 = 60 | 100 × 0.60 = 60 |
| Gross payout | 100 × 1 = 100 | 100 × 0 = 0 |
| Profit or loss before fees | 100 − 60 = 40 profit | 0 − 60 = 60 loss |
The 100 is the gross payout, including the amount originally spent; it is not 100 of profit. A price of 0.60 is not a 60% return either. This example excludes trading fees, network costs, spreads and other costs. Disputed, cancelled or void markets may have different treatment, so the table does not cover every outcome. See the separate guide to prediction-market fees for the costs to check.
Resolution determines which answer applies under the contract. Settlement concerns the resulting payout. An event appearing in the news does not mean the market has resolved; a payout does not mean funds have been withdrawn or received in an external account. The platform and contract rules govern those separate stages.
How does a market probability differ from a poll?
A poll may ask whom you support; a market may ask who will win. You can support one candidate and expect another to win. Even a poll asking whom respondents expect to win produces a different statistical signal from a contract price formed through trading.
A candidate’s contract price described as 60% therefore does not mean 60% of voters support that candidate or predict a 60% vote share. It also does not reveal the views of a whole country or community. Check whether the question concerns the winner, vote share or another condition. Research on prices and beliefs
A market is not a one-person, one-vote survey. People commit different amounts and have different appetites for risk. “Average beliefs” in a research model depends on its assumptions; you cannot automatically treat every displayed price as the arithmetic average of everyone’s opinion.
Four checks before you use a percentage
- Question and rules: Read the event definition, deadline, designated source and resolution conditions. Announcing a policy and bringing it into force are different questions.
- Quote time: A price reflects expectations at a particular time. New information can change it; an old screenshot is not a current quote.
- Trading conditions: Thin participation, a shallow order book or a wide spread can limit what a small trade tells you. Higher volume does not guarantee a correct forecast.
- Other evidence: Read official data, research and event developments alongside the market, checking whether they answer the same question.
Markets can turn dispersed views into observable prices, but information gaps, bias and design limits remain. Evidence that markets were useful in particular historical settings does not establish that every market beats polls, is always accurate or cannot be manipulated. Uses and limitations of prediction markets
How can forecast quality be evaluated?
Assess a series of forecasts, not one winning prediction. Calibration asks whether events assigned similar probabilities happen at corresponding frequencies. Across many comparable forecasts near 70%, look for an occurrence rate near 70%; do not expect every event to happen. Research on calibration (2007)
A Brier score summarizes squared forecast errors. Here we use the binary convention mean((p − y)2): p is the forecast probability from 0 to 1, and y is 1 if the event happens or 0 if it does not. Lower is better, with 0 best. This score does not measure calibration alone. Research on scoring rules (2007)
| Outcome | Squared error |
|---|---|
| YES: y = 1 | (0.70 − 1)2 = 0.09 |
| NO: y = 0 | (0.70 − 0)2 = 0.49 |
For comparisons, use the same questions, probability snapshot times and time remaining before outcomes. Choose the sample and treatment of void or unresolved markets before examining results. Apply the same protocol to a baseline forecast, and disclose limits from small or correlated samples.
A lower score does not guarantee trading profit or platform safety, or show that markets always outperform polls.
Common questions
Does predicting the result correctly guarantee a profit?
No. The entry price, quantity actually filled and fees affect the net result. Getting the direction right and obtaining a favourable trading price are separate matters. The 100-contract payout example is a simplified calculation before costs.
If a 70% event does not happen, was the market useless?
One outcome is not enough to assess a probability forecast. A 70% estimate is not certainty. Assessing forecast quality requires a series of comparable predictions and outcomes, rather than picking a single correct or incorrect result.
Does understanding the mechanism mean I can trade?
No. Reading prices and being eligible to trade are separate questions. Check the particular service’s terms, account requirements and the law that applies to you. An explanation, language option or platform comparison does not grant permission to participate.
Read the question before the number
Keep three things separate: what the contract asks, what the price may reflect and how the outcome pays. If you need the practical differences between orders, fills and settlement, continue with the step-by-step trading guide. Understanding that workflow still does not establish eligibility or legal permission.
For a simulated, noncash learning next step, open HypaForecast’s practice hub.
This guide is published by HypaForecast and is independent of Polymarket. Links to HypaForecast do not establish that you are eligible to use its products or that trading is legally permitted where you are. Check the service’s conditions and applicable local rules separately.
Explore HypaForecast markets
Browse the market list, or inspect the question and current round details for the 5-minute and 15-minute BTC markets.
Sources and scope
Primary academic research informs this guide. The example is hypothetical, uses one unit of account and excludes fees; it is not a live quote or a determination of eligibility or local law.
References for the architecture and forecast-evaluation sections were checked on 17 October 2026. The architecture example uses Polymarket Global’s own documentation, not an independent security audit. This check date applies only to those references.
Share this guide