How prediction-market prices form
Order books vs AMMs
A prediction market has to solve a matching problem before it can solve a pricing problem. Two people who disagree about whether an event will happen need a mechanism that lets them trade against each other, and there are broadly two families of mechanism in use. The first is the central limit order book, familiar from equity and crypto exchanges: traders post bids and asks at prices they choose, and a matching engine pairs compatible orders. The second is the automated market maker, where a pool of capital, usually supplied by liquidity providers, quotes prices algorithmically according to a formula. Both produce a number that looks like a probability, but the number is generated in quite different ways, and the differences matter when you trade.
In an order book, the price you see is the best resting order on each side. If someone is willing to buy a contract paying 1 if the event resolves true at 0.42, and someone else is willing to sell the same contract at 0.44, the market has a bid of 0.42 and an ask of 0.44. Nothing has traded yet. The gap between them, two cents here, is the spread, and it exists because the two sides have not agreed on a price. A trade happens only when someone crosses the spread: either a buyer lifts the 0.44 offer, or a seller hits the 0.42 bid. The book is a queue of intentions, and its shape, how much size sits at each level, tells you how much it would cost to move the price.
An AMM replaces the queue with a formula. The classic constant-product design holds two reserves and maintains the product of their quantities at a constant. In a prediction market context, the two assets are typically the outcome tokens themselves, or a token and its complement. When you buy yes, you add the other asset to the pool and remove yes tokens, and the formula dictates how the ratio shifts. The price is not a resting order; it is the instantaneous slope of the curve, the marginal rate at which the pool will exchange one token for the other right now. There is no counterparty waiting on the other side in the usual sense, only the pool.
The practical consequences diverge in several ways. An order book can be empty or one-sided, especially in a thin market, and then there is no tradeable price at all until someone posts. An AMM is always open: you can buy or sell at any moment, but the price you get depends on how much you trade and how deep the pool is. Order books concentrate liquidity at specific prices and let sophisticated traders place limit orders; AMMs spread liquidity continuously along a curve and remove the need to manage resting orders, at the cost of exposing liquidity providers to adverse selection. Neither is strictly better. It depends on whether the market's problem is finding a counterparty or providing continuous access.
There is also a hybrid pattern worth understanding. Some venues run an order book and route unmatched flow to an AMM, or let market makers quote inside a bonding curve. Others use an AMM as a fallback so that a market never goes dark. When you see a price on such a venue, you may be looking at a book quote, a curve quote, or a blend, and the fee structure can differ depending on which path your order takes. The takeaway is that the same headline number can come from very different plumbing.
For a wallet-first, no-KYC venue like Dexsport, the mechanism tends to be on-chain and transparent: you can inspect the pool or the book before you commit. That transparency is useful precisely because it lets you judge whether the quoted price is a thin, easily moved number or one backed by real depth. It does not tell you whether the market is right, only how it is built.
How buying moves a price
Prices move because someone takes liquidity, and the amount of movement is a function of how much liquidity there is. On an order book, if the lowest ask is 0.44 for 500 contracts and you buy 500, you clear that level and the next ask becomes the best offer. If the next level is 0.46, the market now shows 0.46, even though you paid an average closer to 0.45. Your own order consumed the cheap liquidity, and the remaining sellers are asking more. Buy 5,000 contracts into a book that only has a few hundred at each level and you can walk the price up several cents in a single transaction. This is slippage, and it is not a fee; it is the cost of moving the market against yourself.
An AMM produces the same effect through its curve. With a constant-product pool, the price is the ratio of reserves, and every trade changes that ratio. A small trade barely moves the marginal price; a large trade moves it a lot, and the relationship is nonlinear. Doubling the size of your trade does not double the price impact, it more than doubles it, because the curve steepens as you push the reserves further from balance. This is why AMM quotes are size-dependent in a way that a single order book price is not: there is no single price, only a schedule of prices for different quantities.
It helps to think in terms of the complement. In a binary market, buying yes is economically similar to selling no, because the two tokens sum to 1 at resolution. On an AMM that holds both, a yes purchase drains yes from the pool and adds no, which mechanically raises the price of yes and lowers the price of no. The two prices should stay close to summing to 1, minus fees and any spread, because otherwise there is a free trade. Watching both sides move in opposite directions is a good sanity check that the mechanism is working as intended.
Depth is what determines whether a market is tradeable at scale. A market with a penny spread but only a few hundred dollars at each level looks tight and is actually fragile: one moderately sized order blows through the top levels and the effective price is far worse than the headline. A market with a wider spread but deep resting size may be cheaper for a large trade. The right question is never "what is the price" but "what is the price for my size." Professional traders quote in terms of the average fill and the market impact, not the top-of-book number.
There is a second-order effect that catches people out: your trade changes the information environment. If the market is thin and your order is large, other participants see the move and update their own beliefs. Some may conclude that someone knows something and follow, amplifying the move. Others may decide the price has overshot and fade it, pushing back. On an AMM, the move is mechanical and visible to anyone watching the pool, so it can be interpreted as a signal even when it was just a large uninformed trade. This feedback loop is one reason thin markets can be volatile without any news at all.
Finally, note that moving a price is not the same as changing a probability. If you buy 1,000 contracts and push the quote from 0.40 to 0.45, you have changed what the market will pay, but you have not changed the world. The new price reflects your demand plus the remaining liquidity, not a fresh estimate of the event. A careful reader should treat any price as a snapshot of a fragile equilibrium, one that a single order can disturb.
Price as a live probability
The central claim of a prediction market is that a price can be read as a probability. If a contract pays 1 when an event happens and 0 otherwise, and it trades at 0.60, the market is saying the event is 60 percent likely, in the sense that 0.60 is the break-even probability for a buyer who ignores fees and time value. Buy at 0.60 and you profit if the true chance exceeds 60 percent; sell at 0.60 and you profit if it is below. The price is the point at which the marginal trader is indifferent, and that indifference is what makes it informative.
This reading comes with caveats. The price is a probability only under the assumption that the market is liquid, that participants are roughly risk-neutral at the margin, and that the contract pays exactly 1 or 0. If the payout is anything else, or if there is a chance of the market not resolving cleanly, the arithmetic changes. A contract trading at 0.60 that might pay 0.90 in a disputed resolution is not a 60 percent probability; it is a blend of the event probability and the resolution risk. Careful readers should always check what the contract actually promises before treating the price as a forecast.
Risk preferences distort the mapping too. If the crowd is risk-averse, it may demand a discount to hold a volatile position, so the price sits below the true probability. If the crowd is risk-seeking, or if the contract is being used as a lottery ticket, the price can sit above it. In practice, these distortions are small in deep markets and can be large in thin ones. A market with a handful of participants and no arbitrage capital is not a reliable probability aggregator; it is a small poll with money attached.
Time matters as well. A price of 0.60 with a year to resolution and a price of 0.60 with a day to resolution are different statements, even if the number is identical. The long-dated price embeds more uncertainty about how information will arrive; the short-dated price is closer to a settled fact. Some markets quote a time-decayed probability, others do not, and the difference shows up in how prices drift as resolution approaches. A price that stays flat at 0.60 for months and then jumps to 0.95 in the final hours is not necessarily irrational; it can reflect the slow arrival of confirming information.
The most useful way to read a prediction market price is comparatively. A single number tells you little. Two markets on related questions, or the same market at two points in time, tell you how beliefs are shifting. If a contract on one candidate rises while a contract on a correlated event falls, that is a signal about how the crowd is updating. The absolute level is a probability estimate; the changes are information about the flow of news and the reallocation of capital.
There is also the question of whose probability it is. A prediction market price is the belief of the marginal trader, weighted by capital. If a few large participants dominate, the price reflects their views, not a democratic average. This is not a flaw unique to prediction markets; it is how all markets aggregate. But it means that a price can be well-informed and still unrepresentative, and a reader should ask who is actually trading before treating the number as a consensus.
Arbitrage keeping prices honest
The reason prediction market prices stay coherent is arbitrage. If the same event is tradeable in two places at different prices, or if the yes and no sides of a single market fail to sum to something sensible, someone can lock in a riskless profit by buying the cheap side and selling the expensive one. That activity pushes the prices back together. Arbitrage is not a moral force; it is a mechanical one. The opportunity exists only until enough capital exploits it, and the speed with which it closes tells you how efficient the market is.
The simplest case is the complement. In a binary market, yes and no should sum to about 1, minus fees. If yes trades at 0.55 and no at 0.50, the sum is 1.05, and you can sell both for 1.05 while owing at most 1.00 at resolution, pocketing the difference. Traders will do exactly that, selling yes and selling no until the prices fall back toward a sum of 1. The same logic applies across venues: if yes is 0.55 on one exchange and 0.50 on another, buy the cheap one and sell the expensive one. The residual gap is bounded by the cost of moving funds between venues and the fees on each leg.
Cross-market arbitrage extends to logically related events. If a contract on "candidate A wins" trades at 0.60 and a contract on "candidate A wins the popular vote" trades at 0.65, and it is impossible to win the election without winning the popular vote in the relevant scenario, then the second should not exceed the first. If it does, the arbitrageur buys the first and sells the second. These relationships are not always as clean as they look, because the events may not be nested in the way the prices imply. Careful arbitrageurs check the resolution criteria before assuming a logical link holds.
The limits of arbitrage are real. Capital is finite, and it may be locked in positions or unable to move quickly across chains or venues. Fees, gas costs, and withdrawal delays all eat into the theoretical profit, and a gap that looks like free money on paper can be a loss after costs. There is also the risk that the arbitrage is not actually riskless: the resolution criteria might be ambiguous, one leg might fail to settle, or the venue might halt trading. When a gap persists, it is often because the market is pricing in one of these frictions rather than because everyone is asleep.
Arbitrage also enforces consistency over time. If a price drifts away from where related markets suggest it should be, the gap creates an incentive for someone to trade it back. This is why prediction markets tend to move together in response to shared news, and why a single market rarely stays out of line for long in a liquid environment. The mechanism is not perfect, and it fails in thin or fragmented markets, but it is the main reason a prediction market price is worth reading at all.
A practical implication for traders is that the most reliable edges are often in the relationships between markets rather than in a single price. If you think a market is mispriced, ask what else would have to be true for that to hold. If the answer contradicts another market, one of the two is wrong, and the arbitrage is the trade. If the answer is consistent with everything else, you are making a directional bet against the crowd, which is a different and harder game.
Fees and the spread's role
Every trade pays a toll, and the toll shapes the price you actually get. On an order book, the spread is the most visible cost: you buy at the ask and sell at the bid, and the gap between them is a loss you incur the moment you enter and the moment you exit. If the spread is two cents on a contract trading near 0.50, a round trip costs roughly four cents, or about eight percent of your stake, before any other fees. That is a substantial hurdle, and it is why high-frequency traders care so much about spread and why long-term holders try to enter with limit orders that earn the spread instead of paying it.
AMMs express the same cost differently. Instead of a bid-ask spread, there is a fee on each swap, typically a small percentage, and the price impact of the curve. The fee goes to liquidity providers, who bear the risk of holding the pool's inventory. The impact is the slippage we discussed earlier. For small trades, the fee dominates; for large trades, the impact does. The total cost is the sum, and it can be higher or lower than an order book spread depending on the size and the depth. There is no universal answer to which is cheaper; it depends on the market and the trade.
Fees also create a band within which prices can be "wrong" without being arbitraged. If the round-trip cost is four cents, then a two-cent discrepancy between two venues is not a profit opportunity; it is noise. Arbitrageurs only act when the gap exceeds their costs, so small mispricings can persist indefinitely. This is important for interpreting prices: a market can be off by a couple of cents and still be functioning correctly, because the cost of correcting it is higher than the gain. The size of the no-arbitrage band is a direct function of fees and spreads.
The spread itself carries information. A wide spread signals disagreement, illiquidity, or both. Market makers widen their quotes when they are unsure of the fair value or when they fear being picked off by informed traders. A sudden widening often precedes a news event or reflects a withdrawal of liquidity. A narrow spread, by contrast, suggests confidence and competition among quoters. Watching the spread over time can tell you as much about a market's health as watching the price.
There is a tension between the interests of traders and liquidity providers. Traders want tight spreads and low fees; providers want compensation for the risk they take. In an AMM, that compensation is the fee, and it has to be high enough to attract capital but low enough to keep traders coming. In an order book, it is the spread, and market makers earn it by quoting both sides and managing inventory. If the compensation is too low, liquidity dries up, spreads widen, and the market becomes expensive to use. If it is too high, traders go elsewhere. The equilibrium is set by competition.
For a retail trader, the practical lesson is to count the full cost before trading. The headline price is not the price you pay. Add the spread or the fee, add the slippage for your size, and add any withdrawal or settlement costs. A trade that looks profitable at the quoted price can be a loss after these are included. This is not a reason to avoid prediction markets; it is a reason to trade less often, in larger size when you do, and with limit orders where the venue allows them.
Why prices can still be wrong
A functioning market can still produce a wrong price, and it is worth being precise about why. The first reason is thin liquidity. If a market has few participants and little capital, the price reflects the views of whoever happens to be trading, not a broad consensus. A single large order can move it several cents, and there may be no one willing to take the other side at a better level. In such conditions, the price is a weak signal, and treating it as a reliable probability is a mistake.
The second reason is limited attention. Prediction markets are niche, and many contracts attract little scrutiny. If no one is actively researching the underlying event, the price can drift on stale information or on the trades of a few participants who are not especially well informed. This is not irrationality; it is the natural consequence of a market where the incentive to gather information is small relative to the cost. Deep, high-volume markets attract research; obscure ones do not, and their prices reflect that.
The third reason is that some events are genuinely hard to forecast. A market on a close election or a volatile asset price may be efficiently priced and still be wrong, because the outcome is uncertain. A price of 0.60 does not mean the event will happen 60 percent of the time in any single instance; it means that across many similar situations, the 0.60 bucket should resolve true about 60 percent of the time. Any individual outcome can surprise, and a market that is right on average will still be wrong on many specific questions. Judging a market by a single result is a category error.
The fourth reason is manipulation and distortion. In thin markets, a participant with a strong interest in the outcome can buy contracts to move the price and create a misleading signal, even if they lose money on the trade. This is expensive but not always prohibitively so, and it is more likely in markets with low liquidity and high stakes. Relatedly, some participants may trade for reasons unrelated to the event, such as hedging a position elsewhere or farming a token incentive, and their flow can push the price away from the information-based value.
The fifth reason is resolution risk. A market can be priced correctly for the event and still misprice the contract, because the contract's payout depends on how the resolution criteria are interpreted. Ambiguous wording, disputed facts, or a resolution source that fails can all cause the contract to settle in a way the price did not anticipate. Traders who ignore this risk are effectively buying a probability estimate that excludes a real source of loss. Always read the resolution rules before trading, and treat any ambiguity as a cost.
None of this means prediction markets are useless. It means they are tools with known failure modes, and the price is an input rather than an answer. The most reliable readings come from deep, actively traded markets on well-defined questions, where arbitrage keeps related prices in line and where the cost of being wrong is high enough to attract informed traders. In those conditions, the price is a genuinely useful aggregation of dispersed knowledge. In thin, ambiguous, or obscure markets, it is a number that someone was willing to trade at, and nothing more.
For anyone using these markets as part of a broader betting or trading approach, the discipline is the same as anywhere else: size positions to survive being wrong, count the full cost of trading, and treat any single price as one estimate among many. Prediction markets are not legal advice, and rules vary by jurisdiction; check what applies where you are. If you do trade, keep it 18+ and within a budget you can afford to lose, and use the responsible-gambling tools the venue provides.
This guide was researched, written and fact-checked by the Predict Market Desk editorial desk before publication. Last reviewed 13.10.2026. Spotted something wrong? Tell us.