How to read prediction market prices as probabilities

By Barry Healy EditorPublished: 10 October 2026, 02:44 EDT7 min read

Last reviewed 10 Oct 2026 · Editorial policy

A quiet dusk view of an empty weather station on a windswept hilltop, with a pale gradient sky, a slender glowing ribbon of light rising above the instruments, and a distant city skyline blurred in mist.

What a contract price on Kalshi or Polymarket does & does not tell you about the chance of an event, from displayed midpoints to fees & the favourite-longshot bias.

Price as implied probability

A prediction market contract pays a fixed amount if an event happens & nothing if it does not. Its price can therefore be read as an implied probability, meaning the chance that the price suggests traders collectively assign to the outcome.

Kalshi’s help centre says a contract’s price is “directly tied” to a market-assigned probability running from 0% to 100%. In its own example, a 70% consensus that a company will launch a product means a Yes contract costs 70 cents & a No contract 30 cents.

Kalshi adds that a Yes buyer & a No buyer together contribute $1. As the perceived chance of Yes rises, the Yes price rises & the No price falls.

Polymarket’s documentation says every share is priced between $0.00 & $1.00 & that the price “directly represents the market’s belief in the probability of that outcome”. In its example, a Yes order at $0.60 matches a No order at $0.40 because together they make $1.00.

Those descriptions come from the platforms themselves. Whether a price is an accurate probability is a separate question, covered in the research sections below.

Which price is being quoted?

A market has several prices at once. The bid is the highest price a buyer is offering, the ask is the lowest price a seller will accept, & the gap between them is the spread.

There is also the last traded price & the midpoint, which sits halfway between the bid & the ask. A news report citing “the price” could mean any of these.

Polymarket says its displayed price is the midpoint of the bid-ask spread, unless the spread is wider than $0.10, in which case it shows the last traded price instead.

In Polymarket’s example, a best bid of $0.34 & a best ask of $0.40 produce a displayed price of $0.37. A buyer pays $0.40 & a seller receives $0.34, so the headline figure is not a price anyone trades at in that moment.

Kalshi also runs an order book. According to a University College Dublin working paper, makers post offers stating a side, a price & a quantity, takers match with the most generous offer, & quotes run from 1 cent to 99 cents.

The sources used for this guide do not say which of these figures Kalshi shows as its headline price.

Fees move the break-even point

Fees mean a contract has to win more often than its price implies before a buyer breaks even. The UCD authors note that, before fees, a contract costing X cents must win X percent of the time to break even on average, & that Kalshi’s fees raise that bar.

Kalshi’s fee schedule, marked last updated & effective on 7 July 2026, sets the general trading fee at 0.07 x C x P x (1-P), rounded up, where P is the price in dollars & C is the number of contracts. It applies to orders matched immediately against orders already resting on the book, which are taker orders.

Because P x (1-P) is largest at 50 cents, the dollar fee peaks there. Kalshi’s table lists a fee of $1.75 on 100 contracts at 50 cents, a $50.00 outlay, which works out at about 3.5% of the stake.

At 10 cents, the table lists $0.63 on 100 contracts costing $10.00, about 6.3%. The dollar fee falls towards the extremes, but as a share of the money staked it weighs more heavily on cheap contracts.

Orders that rest on the book & fill later fall under a separate maker formula of 0.0175 x C x P x (1-P), with a multiplier that defaults to 0 unless otherwise indicated. Some named series carry their own multipliers.

Polymarket’s fee page says it charges a taker fee on certain markets, that makers are never charged, & that geopolitical & world events markets are fee-free. The fee in USDC is symmetric around 50%, so a trade at 30 cents costs the same dollar fee as one at 70 cents, & on 100 shares it peaks at $1.75 at 50%.

Polymarket lists separate fee tables for categories such as crypto, sports, & finance, politics, mentions & tech. The rate for each category is not given in the material used here.

Theory: when price equals belief

Reading prices as probabilities is a working assumption, not a given. In NBER Working Paper 12200 (2006), Justin Wolfers & Eric Zitzewitz note that Manski had argued in 2004 that little existing theory supported the practice.

They set out conditions under which prices match the mean belief of traders, & show that for a broad class of models prices are usually close to that average. The key factors, they write, are traders’ degree of risk aversion & the distribution of their beliefs.

Their conclusion is qualified: prices “typically provide useful (albeit sometimes biased) estimates of average beliefs about the probability an event occurs”. An average belief among traders is not the same thing as the true chance of an event.

Evidence: the favourite-longshot bias

A large published test of Kalshi prices comes from Constantin Bürgi, Wanying Deng & Karl Whelan of University College Dublin, summarised in a CEPR VoxEU column on 18 February 2026. Their working paper describes over 300,000 prices from 46,282 contracts across 12,403 events.

They took the final price at close &, where available, earlier prices at 24-hour intervals up to 10 days before close. The column reports that prices broadly reflect outcomes, with a 50 cent contract winning around 50% of the time, & that accuracy improves as markets approach closing.

They also find a favourite-longshot bias. That means cheap contracts, the longshots, win less often than their price suggests, while expensive ones, the favourites, win slightly more often; a 10 cent contract wins less than 10% of the time.

According to the UCD paper, buyers of contracts costing less than 10 cents lose over 60% of their money, while contracts priced above 50 cents earn a small positive return. The average return across all contracts is minus 20% before fees & minus 22% after.

The column reports that takers lose almost 32% on average & makers about 10%. The authors find the bias across politics, entertainment & economic data contracts, & the paper notes some evidence that it is getting smaller over time.

Their model links the pattern partly to people over-estimating small probabilities: it implies that when the true chance is 20%, the average belief is 23%. The authors conclude that Kalshi’s prices “should not be interpreted as unbiased probability estimates”.

Thin markets & manipulation

A price is only as informative as the trading behind it. Polymarket says tighter spreads mean more liquid markets, meaning ones where trades happen readily without moving the price much.

Polymarket also says its order book has no trading size limits but that large orders “may move the price significantly”, & it tells traders to check order book depth, the volume of orders waiting at each price. Where its spread is wider than $0.10, the displayed figure is a single past trade rather than the current offers.

On 22 September 2026 the CFTC’s Division of Market Oversight issued a staff advisory on “mention market” contracts. These settle on whether a person says certain words, attends or appears at an event, or interacts with another person.

The CFTC said such contracts present a heightened risk of manipulation because settlement turns on a person’s conduct that “may be neither independently generated nor externally verifiable”. It reminded designated contract markets, the federally regulated exchanges that include Kalshi, of their duty under Core Principle 3 to list only contracts not readily susceptible to manipulation.

How to quote a market responsibly

On the evidence above, a quoted price carries more meaning when its context comes with it. That means naming the platform & the exact market, & saying which price is meant: midpoint, last trade, bid or ask.

It also means giving the date & time, since prices move, & some sign of how much trading sits behind the figure, such as the spread. A wide spread or thin book signals that the number could shift on a single order.

Wording matters too. Wolfers & Zitzewitz describe prices as estimates of average belief, & the UCD research finds Kalshi prices biased, especially at low prices, so “traders price this at 10 cents” is more precise than “there is a 10% chance”.

The bias research cited here covers Kalshi. These sources do not establish whether the same pattern holds on Polymarket.

Questions

What does a 70 cent price mean on Kalshi?

Kalshi’s help centre says a 70 cent Yes price corresponds to a market-assigned probability of 70%, with the No side at 30 cents. A Yes buyer & a No buyer together put up $1.

Which price does Polymarket display?

Polymarket says it shows the midpoint between the best bid & best ask. If the spread is wider than $0.10, it shows the last traded price instead.

Are prediction market prices accurate?

Research on over 300,000 Kalshi prices found they broadly reflect outcomes & improve near closing, but show a favourite-longshot bias. The authors say Kalshi’s prices should not be read as unbiased probability estimates.

What is the favourite-longshot bias?

It is the pattern in which cheap contracts win less often than their price implies & expensive contracts win slightly more often. On Kalshi, the UCD paper found buyers of contracts under 10 cents lost over 60% of their money.

How do fees change the break-even point?

Fees mean a contract must win more often than its price implies to break even. Kalshi’s taker fee of 0.07 x C x P x (1-P) comes to $1.75 on 100 contracts at 50 cents & $0.63 on 100 contracts at 10 cents.

Sources

Sources checked 9 October 2026. Information only. 18+.

Sections

Alerts

What do you want to understand?