Prediction markets vs sports betting: fees, odds and who actually wins

A sportsbook sells you a price with its margin baked in. A prediction market sells you a contract that pays $1 if you are right and $0 if you are wrong, at a price set by other traders, and charges a fee on top. The two can describe the same game on the same Sunday and cost very different amounts to play. This guide puts the fee formulas next to the bookmaker’s vigorish, works a 50/50 example through both, uses a regulator’s hold data to show what books actually keep, explains how each side settles a result, and ends with where each product is permitted, with sources and without legal advice.

What you are buying

At a sportsbook you stake an amount at odds. At −110 you lay $110 to win $100; the book’s price contains its margin and the book is your counterparty. At a prediction market such as Polymarket you buy shares of an outcome at a price between $0 and $1 from another trader; winning shares redeem for $1.00 and losing shares for $0.00. The platform is not your counterparty; it matches orders and takes a fee. Kalshi, the US exchange, works on the same contract model with its own fee schedule. The price you see is therefore a probability other people are willing to trade at, not a bookmaker’s opinion plus a margin.

The cost of a bet: vigorish and hold

Vigorish is the bookmaker’s commission, built into the odds. On a standard −110/−110 line each side lays $110 to win $100; if the book takes $110 on each side it pays out $210 of the $220 collected and keeps $10. Expressed per unit staked, two-way vig is 100 × (1 − pq/(p + q)) in decimal odds, which gives 4.55% at −110/−110 (decimal 1.909) and 2.38% at −105/−105 (decimal 1.952); the overround on the same −110 line, 110/210 × 2 minus 1, is 4.76%. Different conventions, same money.

Theoretical margin and realised hold are not the same thing. New Jersey’s regulator reported that the state’s sportsbooks held 11.9% of an $816.9 million handle in July 2026 and 6.2% in June 2026. Hold swings with results and with parlay mix, and it is consistently above the two-way vig because parlays and props carry wider margins than a main line. If you want to know what a book keeps, look at hold, not at one quoted line.

The cost of a trade: the fee formulas

Polymarket charges a taker fee only, on certain markets, and makers pay nothing. The formula is fee = C × feeRate × p × (1 − p), where C is the number of shares and p the price. Fee rates by category: crypto 0.07; sports, economics, culture, weather and other 0.05; finance, politics, mentions and tech 0.04; geopolitics 0. Because of the p × (1 − p) term the fee peaks at a 50¢ price and falls toward zero near $0 or $1. There are no deposit or withdrawal fees, and a share of fees is redistributed daily to traders as rebates.

Kalshi’s general formula, as documented by a third-party analysis of its fee schedule, is round up (0.07 × C × P × (1 − P)): 100 contracts at 50¢, a $50 position, costs about $1.75 in fees, and 100 contracts at 10¢ about $0.63. I have not fetched Kalshi’s own schedule PDF, so treat the 0.07 figure as third-party until you check it on kalshi.com.

A worked example at even money

Take a coin-flip game: a 50¢ contract on each side, and a −110/−110 line at the book.

On explicit cost the exchanges are cheaper at even money, and a maker on Polymarket pays nothing. Two caveats keep this from being the end of the argument. First, exchange prices are two-sided: if the best bid is 48¢ and the best ask is 52¢, crossing the spread costs you the 4¢ gap, which on a thin market is larger than any fee and is invisible in the schedule. Second, the book’s 4.55% is on a main line; the realised hold of 6% to 12% in New Jersey tells you what the average bettor actually pays once props and parlays are in the mix. A disciplined main-line bettor pays something near the vig; a casual one pays the hold.

Who sets the price and who limits you

A bookmaker sets and moves its own price and manages risk by limiting accounts. Bovada, to take a crypto-facing US book whose terms I have read, publishes a $0.50 minimum bet, a per-event maximum shown by its “Bet Max” button, a $1,000,000 maximum sportsbook payout, and a clause letting it void wagers and revoke bonuses for players it considers to be abusing limits. On an exchange the price is whatever the order book says, there is no house to limit a winning account, and the constraint is liquidity: you can only buy what someone is selling. Polymarket’s volume gives a sense of scale: $10.57 billion traded in March 2026, about $26.2 billion in the first quarter, with its US sports-only affiliate above $700 million that month. Deep for the headline markets, thin for the long tail.

Settlement versus resolution

A sportsbook grades a bet against the official result and its own house rules; disputes go to support. A prediction market needs a procedure, because nobody is in charge of the answer. Polymarket resolves through UMA’s Optimistic Oracle: once a market ends, a proposer posts the outcome with a bond, $750 per Polymarket’s help centre, which is forfeited if the proposal is wrong; there is a two-hour challenge window; if nobody disputes, the market settles; if someone does, the question escalates to a vote of UMA token holders with a commit and reveal period. Every market publishes its resolution source, end date and edge cases, and the documentation’s own advice is to read the resolution rules before trading. A bet on a game is simple; a contract on “will X happen by date Y” is only as good as its wording, and the fee you saved does not compensate for a resolution you did not read. The Polymarket guide walks through a resolution timeline.

Who wins, by player type

• Main-line bettor, liquid markets. The exchange is cheaper on fees, and cheaper still as a maker; mind the spread.
• Parlay and prop player. The book’s hold on those products is where the 6% to 12% comes from. Exchanges rarely offer the same combinations; the comparison often does not exist.
• Winning bettor. Books limit; exchanges do not. The constraint moves from the operator’s risk desk to the order book’s depth.
• Casual player in a market with a regulated book. The book offers a grading process, a complaints route and a licence; the exchange may be geoblocked for you entirely, which is the next section.

Where each one is permitted

Facts only, dated, and not advice. In the United States the CFTC fined Polymarket’s operator $1.4 million in January 2022 for running an unregistered binary options facility and ordered a wind-down; polymarket.com’s terms block US trading and point to a separate US entity. Kalshi, a CFTC-registered exchange, won in federal court in September 2024 against the CFTC’s ban on its congressional contracts, the appeals court denied a stay in October 2024, and in May 2025 the CFTC moved to dismiss its own appeal. On 4 February 2026 the CFTC withdrew its 2024 proposed rule on event contracts and a related staff letter, announcing new rulemaking to follow. State-level actions against prediction markets are ongoing and I have not verified them individually.

Outside the US, Polymarket’s own geoblocking page lists close-only status for Australia, Belgium, Brazil, parts of Canada, France, Germany, Italy, New Zealand, Poland, Singapore, Slovakia, Taiwan, Thailand, the UK and the US, among others. The UK Gambling Commission said in February 2026 that prediction markets could not be classified as non-gambling and would need a betting intermediary licence. France’s regulator ordered Polymarket blocked in July 2026 after banning transactions to it in November 2024; Belgium’s gaming commission blacklisted it in January 2025. Sportsbooks, crypto or otherwise, publish their own restricted lists in their terms. The country-by-country guide keeps the table; the prediction markets category lists the platforms I have reviewed, and the crypto sportsbooks category the books.

The short version

A book charges a margin inside the price, about 4.55% on a two-way main line and 6% to 12% of handle in practice. An exchange charges a fee on top of a market price, about 1.25% of notional on a Polymarket sports market at 50¢ for a taker and nothing for a maker, plus whatever the spread costs you. Exchanges are cheaper for liquid, simple markets and for winners; books are the only option for parlays and for most people in most regulated countries, where the exchanges are blocked. Read the resolution rules on one and the house rules on the other, and read the Bovada review for what a crypto-facing book’s limits look like in practice.