Order Books

Every price on a prediction market looks like someone's official verdict. The screen shows Yes at 34¢ and No at 66¢, yet no one at the venue chose those numbers, and there is no bookmaker adjusting them from a back office. The numbers come from an order book, a running list of what traders are currently willing to pay and accept, and the mechanics of that list answer most of the questions newcomers bring to these markets: who decides the price, who is on the other side of a trade, and why the number moves the moment news breaks.

Who sets the odds when there is no bookmaker?

At a sportsbook, the odds are set by the operator. Its trading team publishes a line, moves it as money arrives, and takes the other side of every position its customers open. The operator is the counterparty, which is why sportsbook customers call it "the house." Customer losses are its revenue, and the posted line carries a built-in margin to keep that revenue reliably positive.

A prediction market has no equivalent role, because its odds are not set at all. They are prices, formed by people trading, the way prices form in any open market. The venue lists an event contract, an instrument that settles at $1 if a defined event occurs and $0 if it does not (the CFTC publishes a plain-language explainer of how these contracts work), and then operates a marketplace where anyone can submit orders to buy or sell either side. The differences follow from that design:

The price you see on an exchange is not anyone's verdict. It is the current meeting point between what buyers will pay and what sellers will accept, recalculated continuously as orders arrive. Polymarket's help center describes its prices as a function of real-time supply and demand, which is an accurate description of any order book market. Because each contract settles at $1 or $0, that meeting point doubles as a probability estimate. A 34¢ price implies roughly a 34% chance, a conversion covered in Prices and Probabilities. The broader comparison between the two models has its own guide, Markets vs Sportsbooks.

How does a prediction market order book match buyers and sellers?

An order book is two lists. Bids are buy orders, each naming a price a trader is willing to pay and a quantity. Asks are sell orders, each naming a price a holder is willing to accept and a quantity. The highest bid and the lowest ask are the two numbers that matter most, and the gap between them is called the spread. Suppose the book for a Yes contract looks like this:

Here the best ask is 35¢, the best bid is 33¢, and the spread is 2¢. Venues typically display the midpoint of those two numbers as the market price, 34¢ in this case. Polymarket's documentation describes exactly this convention.

Orders arrive in two basic forms. A limit order names a price, say 33¢ or better to buy 200 contracts, and waits. A market order accepts whatever the book offers right now. When a new order reaches the exchange, the matching engine checks whether it crosses the best price on the opposite side. If it does, it fills immediately against the resting orders, best price first; among orders at the same price, exchanges typically fill the earliest one first. If it does not cross, it joins the book and waits.

A worked example shows why size matters. A market order to buy 300 contracts in the book above takes all 200 available at 35¢, then 100 of the 300 available at 36¢, for an average price just above 35.3¢. The 35¢ level is now empty, the best ask is 36¢, and the displayed midpoint has moved from 34¢ to 34.5¢. Nothing about the event changed; one order consumed the cheapest available supply.

This design, a central limit order book, is the standard mechanism on the major prediction market venues. Some venues use an automated market maker instead, an algorithm that prices every trade against a pooled inventory. Research from Paradigm examines how such designs behave for event contracts. The trade-offs between the two approaches are covered in Venue Types.

Who is on the other side of a prediction market trade?

The other side of a trade is always another participant, never the venue. Polymarket answers the question directly in its help center: it is not the house, and every trade is matched peer to peer. Any exchange-model venue works the same way. The operator matches orders between customers rather than trading against them, standing between the two sides the way a stock exchange stands between two shareholders without owning the shares.

Mechanically, the other side of a Yes position is a No position. When one trader offers 34¢ for Yes and another offers 66¢ for No on the same market, the two payments add up to exactly $1, and the exchange matches them by creating one contract of each, a Yes for the first trader and a No for the second. New contracts come into existence this way, funded entirely by the traders on the two sides. Matching also happens by transfer when an existing holder sells to a new buyer directly. In that case no new contracts are created, and the position simply changes hands.

$$P_{\text{Yes}} + P_{\text{No}} \approx \$1$$

The relationship holds because the pair settles for exactly $1 between them, the full dollar going to one side at resolution and nothing to the other. Prices track that identity closely. When Yes trades at 34¢, No tends to trade near 66¢, because any persistent gap between the pair and $1 is money another trader can collect by buying both sides.

Knowing that the counterparty is a fellow participant still leaves the practical question of who that participant tends to be. Three profiles cover most of the book:

Another individual trader who reads the event differently. Two people can look at the same election, match, or product launch and reach opposite conclusions, and the market exists precisely to let those views transact against each other.

A market maker whose resting quote your order matched. Market makers place standing orders on both sides of a market to earn the spread rather than to express a view on the outcome. The next section examines how they work.

A better-informed participant. Some traders bring research, domain expertise, or faster reactions to news. A working paper analyzing Polymarket's transaction history attributes much of the market's forecasting accuracy to a small minority of persistently skilled traders, on the order of a few percent of participants, whose profits are funded by the rest of the market's losses. The finding carries the usual working-paper caveats, but the structural point is durable: some share of the flow on the other side is professional.

At the level of a single contract pair, the arithmetic is zero-sum before fees. The dollar the two sides put in is the same dollar one side takes out at settlement. This is what trading against peers rather than against the house means in practice. The market redistributes money between participants according to who was right, and the exchange stands outside that transfer.

What do market makers do in prediction markets?

A market maker is a participant who quotes both sides of the same market at once: a standing bid below the current price and a standing ask above it. A maker might bid 33¢ and offer at 35¢ simultaneously. When a seller hits the bid and a buyer later lifts the offer, the maker has bought at 33¢ and sold at 35¢, collecting the 2¢ spread without needing an opinion on whether the event happens. Repeated across thousands of contracts and many markets, that spread is the business.

Exchanges depend on this activity. An order book with no resting orders has nothing for a newcomer to trade against; every arriving order would wait for a stranger to show up wanting the exact opposite position at the exact same moment. Makers keep the book populated so that anyone can trade immediately at a known price, and competition among makers is what pulls spreads tight. In thin markets with little maker activity, spreads widen and every entry and exit costs more.

Market making carries two risks that shape how makers behave. The first is inventory risk: fills tend to arrive unevenly, so a maker who keeps buying from sellers accumulates a growing Yes position, and if news breaks against it the inventory loses value faster than spread income accrues. The second is adverse selection: the traders most likely to hit a resting quote in the seconds after news are the ones who saw the news first, so stale quotes are systematically picked off at bad prices. Both pressures push makers to widen their quotes or withdraw entirely around scheduled announcements and fast-moving stories, which is why spreads often widen sharply at exactly the moments trading interest peaks.

This structure produces the maker and taker vocabulary that appears throughout exchange documentation. An order that rests in the book provides liquidity and is a maker order; an order that fills immediately removes liquidity and is a taker order.

Venues tend to price the two roles differently, charging takers while rewarding makers through rebate or incentive programs such as Polymarket's liquidity rewards. An independent analysis of roughly 72 million Kalshi trades put a number on the economic gap between the roles, estimating that takers earned about one percentage point less per trade than a neutral benchmark, with a mirror-image edge for the makers they traded against, and with the size of the transfer varying widely by market category. Because it is a single analysis, its magnitudes deserve caution. The direction is what the structure predicts: immediacy is a service, and takers pay for it.

How do prediction market exchanges make money?

A sportsbook's revenue is inseparable from its customers' results. It holds the other side of every position its customers take, so what customers lose is, in aggregate, what the operator earns, and the margin inside its odds exists to keep that aggregate dependably positive. An exchange has no such exposure. It cannot win or lose on an outcome because it holds neither side of any contract.

What an exchange sells is the marketplace itself, and it charges for use the way other marketplaces do, through fees on activity. Kalshi states its model plainly: revenue comes from transaction fees, and the exchange has no financial stake in which way outcomes swing. Fee designs differ from venue to venue and change over time, so current schedules are best read at the source: Polymarket documents its fees here and Kalshi documents its fees here.

The distinction shapes incentives. A business paid per trade benefits from more trading: more markets, more participants, tighter spreads, more reasons to transact. Whether any particular trader wins or loses tends not to move the operator's revenue at all. The operator's interest sits with the health of the marketplace, its volume, liquidity, and participation, rather than with the results of the people trading on it.

How does trading move prediction market prices?

The displayed price is derived from the order book, so it moves whenever the book moves. Two mechanisms account for nearly every price change.

Trades consume depth. In the earlier example, a single market order to buy 300 contracts cleared the 35¢ level, pushed the best ask to 36¢, and moved the displayed price from 34¢ to 34.5¢ in one step. Larger orders walk further up the ladder and move the price more. This is price impact, and it scales with the size of the order relative to the depth resting in the book.

Quotes move without trades. The book also changes when traders cancel and replace their resting orders. When a headline lands, makers and other participants reprice their quotes within seconds, often before a single trade prints. A market price can jump on news with no volume at all, because the people willing to buy and sell simply changed their terms.

Beneath the mechanics sits the reason these markets are worth watching at all. Every order in the book is an opinion about the outcome with money attached, and the matching process nets those opinions into a single number continuously. When the number moves, some set of participants has changed what they are willing to pay, usually because information arrived: a poll, an injury report, a court filing, an earnings release. Economists have studied this aggregation for decades. The field's foundational survey by Wolfers and Zitzewitz documents how prediction market prices behave as probability forecasts, drawing on information scattered across many participants that no single forecaster holds. The logic of why many funded opinions tend to outperform any individual expert is the subject of Wisdom of Crowds.

That is the full answer to the question this guide opened with. No single participant sets the price. Every order placed, filled, canceled, or repriced adjusts it, and the order book is where all of those adjustments meet. The result is a number that nobody controls and everybody contributes to, updated as fast as its participants can change their minds.

Related guides