Prediction Markets Explained: How an Event Contract Works

By · author of the MIT-licensed teachersbettextbook library ·

No line, no juice, nobody on the other side who wants you to lose. Just a stack of offers from other people, and a number that is already a probability.

Meet your classmates

Same five, and today Brian is out of a job.

A contract is a dollar bill that might not exist

It pays $1 if the thing happens. It pays $0 if it does not. That is the whole product.

So if Chalk to win is trading at 62 cents, you are paying 62 cents for a shot at a dollar. Nothing to convert. Nothing to decode.

Which means the price is already a probability. 62 cents is 62%. Bookie Brian writes the same idea as −163 and leaves you to dig it back out.

Nobody here writes the price

This is the part that catches people. Brian is still in the room. He is just not setting the number any more.

What he runs is a list. Everyone willing to buy, everyone willing to sell, and how many contracts each of them wants. That list is the order book, and the list is the price. He charges a fee for keeping the room open. He is not on either side, so he does not care who wins.

Joe hates this. He wants one number, posted, so he can bet it. There is no one number. There are two, and a gap between them.

The middle of the gap is the lazy answer

Buyers are stacked up at 61 cents, 400 contracts deep. Sellers are at 63 cents, and there are only 100 of them. Split the difference and you get 62. That is the midpoint, and it is the number almost every screenshot quotes.

The two sides are not the same size though. One is a wall and one is a screen door. Weight the price by how much is standing on each side and it comes out at 62.6, leaning at the thin side, because the thin side is the one about to go. That weighted number has a name. It is the microprice.

Then somebody buys 400.

Market: Chalk to win each contract pays $1 ASKS: what sellers will take BIDS: what buyers will pay sizes in contracts 30066¢ 18065¢ 12064¢ 10063¢ 40061¢ 35060¢ 50059¢ the spread midpoint 62.0¢ microprice 62.6¢ buy 400 new midpoint 63.5¢ average fill 64.2¢ Average Joe wants a number to bet Bookie Brian takes a fee, not a side +EV Evan reads both stacks Nobody wrote this price. It is a list of people. 400 buyers at 61¢. Only 100 sellers at 63¢. The midpoint says 62.0. The thin side says 62.6. One order took every seller under 66¢.
Not a simulation this time, because nothing here is random. One book, one order, and arithmetic. Bids grow left, asks grow right, and the empty band between them is the spread. 400 contracts are bid at 61¢ and only 100 are offered at 63¢. The midpoint of that gap is 62.0¢. Weighted by what is standing on each side it is 62.6¢, and the lean is toward the sellers because there are barely any. Then one buy for 400 clears them out at an average of 64.2¢ and the next seller wants 66¢, so the book reopens at a midpoint of 63.5¢. The midpoint was not so much wrong as not looking.

A sportsbook sells you a price it wrote. An exchange shows you what other people are willing to do, and charges you to join in. Those are different products that happen to look alike.

What it costs you

Two costs, and you can work out both before you click. The first is the spread you cross: on a 61 cent bid and a 63 cent ask, buying right now costs a cent over the midpoint. The second is the fee, which Kalshi publishes.

taker fee=⌈ 0.07 C P (1−P) ⌉cent\text{taker fee} = \left\lceil\, 0.07\,C\,P\,(1-P) \,\right\rceil_{\text{cent}}

P(1−P)P(1-P) peaks at a coin flip, so the fee tops out at 1.75¢ per contract at 50¢ and falls toward the tails. That is the exact inverse of sportsbook vig, which lands heaviest on longshots. Because the fee is charged per trade and settlement is free, a round trip pays it twice while holding to resolution pays it once.

Worked example: 100 contracts at 62¢

outlay=100×$0.62=$62.00\text{outlay} = 100 \times \$0.62 = \$62.00
fee=⌈ 0.07×100×0.62×0.38 ⌉=⌈$1.6492⌉=$1.65\text{fee} = \lceil\, 0.07 \times 100 \times 0.62 \times 0.38 \,\rceil = \lceil \$1.6492 \rceil = \$1.65
all-in=$63.65  ⟹  pbreak-even=63.65%\text{all-in} = \$63.65 \;\Longrightarrow\; p_{\text{break-even}} = 63.65\%
If your own fair estimate is p=66%:EV=0.66×$100−$63.65=+$2.35\text{If your own fair estimate is } p = 66\%:\quad \mathrm{EV} = 0.66 \times \$100 - \$63.65 = +\$2.35
Two numbers that must never be merged: the fair probability (66%, your estimate) and the fee-adjusted break-even (63.65%, arithmetic on the posted price). The gap between them is the whole calculation. Put your estimate at 63.65% instead and the EV is $0. That is the honest answer, not a rounding opportunity.

Run it on your own price: the prediction market odds converter turns a contract price into American, decimal, fractional and implied probability, and prints the fee-adjusted break-even beside it.

What is settled, and what is not

Settled. Kalshi is a designated contract market regulated by the CFTC under the Commodity Exchange Act. Its contracts are financial instruments on a federally regulated exchange, not wagers taken by a licensed gaming operator. That is a genuinely different legal category from a sportsbook, and it is why an account there opens under securities-style identity rules rather than a state gaming licence.

Not settled, and we are not going to pretend otherwise:

Content farms in this niche state all three as though they were resolved. They are not. This page is information about market mechanics, not legal or tax advice, and none of it suggests you open an account or take any position.

Want more math?the Nerd Corner

Nerd Corner

Advanced material. Nothing above depends on it.

An order book, not a posted line

The structural difference from a sportsbook is who is on the other side of you. A book is your counterparty: it writes the price, holds the opposite position, and pads both sides so it profits on balanced action. An exchange holds no position ever. It publishes two stacks of resting orders, bids (what buyers will pay) and asks (what sellers will take), matches a YES against a NO, and charges a fee either way.

Two roles fall out of that. A maker rests an order and waits to be filled. A taker crosses the book and trades now, at the posted price. The taker pays for immediacy; the maker gets paid for supplying it. There is no equivalent choice at a sportsbook, where every bet is a take.

Three consequences worth spelling out:

  • The cost is explicit. A book’s margin is hidden inside the price. An exchange charges a stated fee and shows you the spread you are crossing. Both are costs; only one is itemised.
  • Depth is visible. A posted line tells you nothing about how much money stands behind it. An order book tells you exactly how many contracts sit at each price, and therefore how far your own order will move it.
  • Winning accounts aren’t a liability. A book’s profit depends on you losing, which is why books limit or ban consistent winners. An exchange earns the same fee whichever side wins.

How a book gets read properly (bid, ask, depth, and why the midpoint is the lazy answer) is its own page.

Why the price is the probability

The body says 62 cents is 62%. That is an identity, not a slogan, and here is where it comes from.

A contract has exactly two payoffs. There is nothing in between:

payoff={$1.00if it resolves YES$0.00if it resolves NO\text{payoff} = \begin{cases} \$1.00 & \text{if it resolves YES} \\[2pt] \$0.00 & \text{if it resolves NO} \end{cases}

Call the event’s real chance pp. The expected payoff is what the contract returns on average, which is one line of arithmetic:

E[payoff]=p×$1+(1−p)×$0=p\mathbb{E}[\text{payoff}] = p \times \$1 + (1-p) \times \$0 = p

So the average payoff is the probability, in dollars. A price below pp returns more than it costs on average and gets bought. A price above it gets sold. The number the crowd stops at is their estimate of pp, wearing a dollar sign.

P=0.62  ⟺  62% impliedPNO=1−PYESP = 0.62 \;\Longleftrightarrow\; 62\%\ \text{implied} \qquad P_{\text{NO}} = 1 - P_{\text{YES}}

One honest caveat. That derivation assumes a trader who cares only about the average and not about the risk itself. Real people care about both, which is one reason the tails lean. There is a section on that further down.

The second identity is the one a sportsbook price cannot manage. A single exchange price sums to 100% by construction: YES at 62 cents and NO at 38 cents are the same claim read from opposite ends. A book’s two prices sum past 100%, and you have to de-vig them before either one means anything.

An exchange has an overround too. It is just the spread

One price sums to 100%. A two-sided quote does not, and the amount it overshoots by is not a coincidence.

Take the book in the figure above: 61 bid, 63 ask on Chalk. Buying YES right now costs 63 cents. Buying NO right now costs whatever NO is asking, and NO’s ask is the mirror of the YES bid:

askNO=100¢−bidYES=100−61=39¢\text{ask}_{\text{NO}} = 100\text{¢} - \text{bid}_{\text{YES}} = 100 - 61 = 39\text{¢}

Buy both sides and you pay 63 + 39 = 102 cents for a pair that is certain to return exactly $1, because one of them settles at a dollar and the other at zero. Two cents over, on a two-cent spread. Those are always the same number:

m=askYES+askNO−100¢=askYES−bidYES=spreadm = \text{ask}_{\text{YES}} + \text{ask}_{\text{NO}} - 100\text{¢} = \text{ask}_{\text{YES}} - \text{bid}_{\text{YES}} = \text{spread}

An exchange’s two-way margin is its spread, to the cent, before fees. A sportsbook has the same overshoot and calls it the vig, but it is welded into the two prices and you have to do arithmetic to find it. Here it is the gap you can already see on the screen.

The microprice, worked

The body says the size-weighted price comes out at 62.6 cents. Here is where that came from.

The midpoint ignores size. It averages the two best prices and stops:

mid=bid+ask2=61+632=62.0¢\text{mid} = \frac{\text{bid} + \text{ask}}{2} = \frac{61 + 63}{2} = 62.0\text{¢}

The microprice weights each price by the size standing on the other side. That looks backwards for about ten seconds and then it clicks: a huge bid stack means plenty of people want in, so the next trade is more likely to happen up at the ask.

microprice=bid×Qask+ask×QbidQbid+Qask\text{microprice} = \frac{\text{bid} \times Q_{\text{ask}} + \text{ask} \times Q_{\text{bid}}}{Q_{\text{bid}} + Q_{\text{ask}}}
=61×100+63×400400+100=31,300500=62.6¢= \frac{61 \times 100 + 63 \times 400}{400 + 100} = \frac{31{,}300}{500} = 62.6\text{¢}

It leans at the thin side, because the thin side is the one about to go. The sanity check is what happens when both stacks match: set Qbid=QaskQ_{\text{bid}} = Q_{\text{ask}} and the weights cancel, and the formula collapses back to the plain midpoint.

Where the 64.2 cent fill came from

An order for 400 contracts cannot all fill at 63 cents, because only 100 are offered there. It eats the stack a level at a time and pays more at each one:

100 @ 63¢  +  120 @ 64¢  +  180 @ 65¢  =  400 contracts100 \text{ @ } 63\text{¢} \;+\; 120 \text{ @ } 64\text{¢} \;+\; 180 \text{ @ } 65\text{¢} \;=\; 400 \text{ contracts}
cost=6,300+7,680+11,700=25,680¢25,680400=64.2¢\text{cost} = 6{,}300 + 7{,}680 + 11{,}700 = 25{,}680\text{¢} \qquad \frac{25{,}680}{400} = 64.2\text{¢}

The posted ask said 63. The order paid 64.2. That gap is slippage, and it is why a top-of-book quote is not a price for any size you feel like. Afterwards the best remaining ask is 66 against a 61 bid, so the quoted midpoint jumps to 63.5 without anybody changing their mind. Depth moved it, not news.

The venues, and how they differ

“Prediction market” covers three quite different machines. Status below is a snapshot dated 24 July 2026; access rules and fee schedules are set by the venues and change without notice.

VenueStructureSettles byUS access
KalshiCentral limit order book, US dollars, CFTC-regulated designated contract marketExchange, against the contract’s written rulebookOpen to US users; contested by some state regulators
PolymarketCentral limit order book on-chain, USDC collateralUMA optimistic oracle: proposal, challenge window, token-holder voteHistorically restricted for US persons; status has moved, so verify
Betfair ExchangeBack/lay exchange in decimal odds, commission on net market winningsExchange, against published market rulesNot available to US users

Betfair is worth one extra sentence because its quoting convention hides the same object. “Backing at 2.50” and “buying at 40¢” are the same trade: decimal odds dd invert to an implied probability, and that probability is the contract price.

P=1dd=2.50  ⇒  P=0.40  ⇒  40¢P = \frac{1}{d} \qquad d = 2.50 \;\Rightarrow\; P = 0.40 \;\Rightarrow\; 40\text{¢}

Kalshi is the only one of the three that this site ingests. Polymarket and Betfair data sit behind a legal review before anything from them is displayed to US readers, so the numbers on our board come from Kalshi and from licensed sportsbook feeds, never from scraping any venue.

What a price is not

A prediction-market price is an estimate produced by whoever showed up, weighted by how much money they brought. It is often a very good estimate. It is not an oracle, and two failure modes are measurable.

The tails lean. A 2026 study of more than 300,000 settled Kalshi contracts found prices informative and sharpening toward close, and low-priced contracts winning less often than the fee-adjusted break-even required.[2] That is the favorite–longshot bias, first measured at racetracks in 1949[3] and found in essentially every betting pool since.[4] Our fair-line machinery shades longshots down for exactly this reason; treating an exchange as unbiased is a modelling error with a price tag.

Quiet books lie. A tight-looking spread on a market nobody has touched in three hours is decoration. Postponed games are the sharpest version: the market stays open, the resting orders go stale, and the quote becomes a fossil that still renders like a live price. Every number on this site carries both timestamps (when we computed it and when the venue last moved it) because staleness is the failure no screenshot reveals.

Open the free prediction market odds converter →Contract price to American, decimal, fractional and implied probability, plus the fee-adjusted break-even most converters skip.

Sources

  1. Kalshi. “Fee Schedule” (July 2026 revision). The taker/maker formulas, the round-up-to-the-cent rule, and the zero settlement fee.
  2. Burgi, C., Deng, W. & Whelan, K. (2026). “Makers and Takers: The Economics of the Kalshi Prediction Market.” GWU Working Paper 2026-001. PDF. 300k+ settled contracts: informative prices, favorite–longshot bias after fees.
  3. Griffith, R. M. (1949). “Odds Adjustments by American Horse-Race Bettors.” American Journal of Psychology 62(2), 290–294.
  4. Thaler, R. H. & Ziemba, W. T. (1988). “Anomalies: Parimutuel Betting Markets: Racetracks and Lotteries.” Journal of Economic Perspectives 2(2), 161–174.
  5. Wolfers, J. & Zitzewitz, E. (2004). “Prediction Markets.” Journal of Economic Perspectives 18(2), 107–126. The standard survey of why a contract price reads as a probability.

Every formula here lives on the Formula Sheet for quick reference.

Check your understanding

Frequently asked questions

What is a prediction market?

A prediction market is an exchange where people trade binary event contracts. Each contract pays $1 if a stated event happens and $0 if it does not, so its trading price in cents is the market's implied probability of that event. The exchange matches buyers against sellers and never takes a position.

What is an event contract?

An event contract is a claim on a stated, verifiable outcome that settles at $1 if the outcome occurs and $0 if it does not. Buying YES at 62¢ costs $0.62 per contract and returns $1.00 if the event resolves YES. Buying NO at 38¢ is the same trade from the other side.

How is a prediction market different from a sportsbook?

A sportsbook posts a price, takes the other side of your bet, and pads both sides so the implied probabilities sum above 100%. A prediction market runs an order book: your counterparty is another trader, the venue charges an explicit fee instead of an overround, and your cost is the spread you cross plus that fee.

Who regulates prediction markets in the United States?

Kalshi operates as a CFTC-regulated designated contract market, so its event contracts sit under federal commodities law rather than state gaming law. Whether sports event contracts also fall under individual states' gaming statutes has been contested in court and is unsettled. This page is information, not legal advice.

Does a contract price tell you the true probability?

It is an estimate, not an oracle. Research on 300,000+ settled Kalshi contracts found prices informative and sharpening toward close, yet low-priced longshot contracts won less often than the fee-adjusted break-even required. Treat a price as one estimate among several, and shade the tails.