Prediction Markets and Sportsbooks: Comparing the Price You Pay
At its 0.05 sports rate, Polymarket's fee takes 4 percent of a stake at 0.20 and 1 percent at 0.80.
A sportsbook charges through the gap between its two odds; a prediction market charges a taker fee when you buy, plus any spread between its Yes and No prices.
Put both on one scale, expected cost per unit staked, and the order flips with the prices: odds of 1.95 and 1.95 against 0.50 and 0.50 cost -2.50 and -2.44 percent, while 1.80 and 2.10 against 0.56 and 0.46 cost -3.08 and -5.92 percent. The tool below does that arithmetic on your own prices. Whether you may trade on a prediction market at all depends on where you live.
- Source
- Polymarket's published fee documentation
- Version
- The fee page carries no date or version
- Limit
- Taker fee at the 0.05 sports rate, before any rebate
Where does each side take its cut?
A sportsbook's cost is its margin: two odds whose implied probabilities add up to more than 1, so the house keeps a slice of the money staked whichever side wins. How a Pinnacle margin is read has the formula; here only its result matters, a percentage of the stake. The wider prediction markets hub explains which operators exist and who can reach them.
A prediction market has no odds to shave. A share pays 1 if the outcome happens and trades at a price between 0 and 1, which is the implied probability itself. The cost arrives as a fee taken at purchase and as a spread: each side of the same event is quoted on its own, and when the two buy prices add up to more than 1.00, the excess is paid on top of any fee. The fee formula leaves that excess out.
Polymarket's documentation gives the taker fee as shares times rate times price times one minus price. The rate listed for sports is 0.05, set market by market, so a given market can carry fees switched off or another rate (the fee page does not date itself). Makers, whose orders rest in the book until someone takes them, are never charged, and Polymarket says it shares part of the taker fees with them, 15 percent in sports. A taker rebate program also exists and its amounts could not be read, so an active trader may pay less.
Rounding has a quirk worth knowing if you ever size a very small order. Fees are rounded to five decimals and anything below the smallest unit becomes zero, so trades near the extremes can carry no fee at all, although the spread still applies.
What about Kalshi?
Kalshi's help page says it charges a transaction fee on the expected earnings of a contract, that some markets carry other fees, and that resting orders which fill can pay a maker fee. Its full fee schedule could not be read, so this page quotes no Kalshi figure and sets up no comparison.
How do prediction market fees change with the price?
The formula multiplies by the price and by one minus the price, so in dollars the fee peaks at 0.50 and shrinks toward either end. On 100 shares at 0.50 it is 1.25, the largest dollar fee the formula gives on 100 shares at the sports rate, and 2.5 percent of the 50 staked. The curve below shows the same fee on two scales.
Taker fee by share price, two ways of reading it
CalculatedShow the data as a table
| Price per share, from 0 to 1 | Fee as a percentage of the stake | Fee as a percentage of the gross gain |
|---|---|---|
| 0.05 | 4.75% | 0.25% |
| 0.10 | 4.50% | 0.50% |
| 0.15 | 4.25% | 0.75% |
| 0.20 | 4.00% | 1.00% |
| 0.25 | 3.75% | 1.25% |
| 0.30 | 3.50% | 1.50% |
| 0.35 | 3.25% | 1.75% |
| 0.40 | 3.00% | 2.00% |
| 0.45 | 2.75% | 2.25% |
| 0.50 | 2.50% | 2.50% |
| 0.55 | 2.25% | 2.75% |
| 0.60 | 2.00% | 3.00% |
| 0.65 | 1.75% | 3.25% |
| 0.70 | 1.50% | 3.50% |
| 0.75 | 1.25% | 3.75% |
| 0.80 | 1.00% | 4.00% |
| 0.85 | 0.75% | 4.25% |
| 0.90 | 0.50% | 4.50% |
| 0.95 | 0.25% | 4.75% |
Read in percent of the stake, the fee falls as the price rises: 4.5 percent for 100 shares bought at 0.10, 0.5 percent at 0.90 (the dollar fee is 0.45 both times, on stakes of 10 and 90). Read in percent of the gross gain, it climbs the other way. The fee is taken at purchase whether the share wins or loses, so the stake view is the fair one next to a sportsbook margin, which is also a share of the stake. Buying the longshot costs the most per dollar staked and buying the favourite the least.
Can you run the comparison with your own prices?
Type the two decimal odds of a two-way market from a sportsbook, then the two prices you would pay to buy each side of the same event on the prediction market. Buy prices means asks: a last trade or a midpoint is not what you would pay, and a bid is what someone pays you. The fields start empty, apart from the fee rate, which starts at 0.05; change it if your market shows another rate, or enter 0 for a market without fees.
Prices move between the moment you copy them and the moment you order. Copy both sides together and note the time next to each number, or the comparison quietly mixes two different hours.
Fee and margin side by side
Every price in the worked cases below is invented: illustrative prices, not quotes. Yours go in the fields.
Reads only the numbers typed above. The figures stay in your browser.
The result lists the sportsbook margin on your pair, the chance of your selection once that margin is removed, the fee per share, the sum of the two buy prices, and the expected result per unit staked in three forms: at the sportsbook, on the market paying the fee, and on the market with a resting order that fills and pays none. If the two buy prices add up to less than 1.00 the tool stops, because those are almost certainly sell prices or a stale book, and the calculation would show a profit that does not exist. The margin is removed proportionally (each implied probability divided by the sum), and the chance that results is treated as the true one.
The comparison is about cost only. It leaves out:
- Who is right about the event.
- Slippage on a large order.
- Settlement rules and any promotion.
- Any margin removal other than proportional.
Three cases, worked with invented prices
Case A: sportsbook odds of 1.95 and 1.95 against Yes and No both at 0.50. The margin is 2.56 percent and each side has a fair chance of 50 percent. The expected result per unit staked is -2.50 percent at the sportsbook and -2.44 percent on the market (a fee of 0.0125 per share at 0.50). The two land 0.06 points apart, close enough that a different removal method could reverse them. A resting order that fills pays no fee and shows 0.00 percent, but it may never fill.
Case B: odds of 1.80 and 2.10 against Yes at 0.56 and No at 0.46, compared on the Yes side. The margin is 3.17 percent, the fair chance of Yes 53.85 percent, and the two buy prices add up to 1.02, a spread of 2.00 percent. The fee is 0.01232 per share, 2.20 percent of the price. Result: -3.08 percent at the sportsbook, -5.92 percent on the market. The resting order still shows -3.85 percent, because the Yes price sits above the fair chance.
Case C: odds of 1.30 and 3.80 against Yes at 0.75 and No at 0.27. The margin is 3.24 percent and the favourite's fair chance 74.51 percent. On the favourite the fee is 0.00938 per share, 1.25 percent of the price: -3.14 percent at the sportsbook, -1.88 percent on the market. On the outsider, 3.80 against No at 0.27, the fee is 3.65 percent of the price and the market shows -8.92 percent against the same -3.14 at the sportsbook.
Across the three cases the market is level, worse, then better on one side and far worse on the other. A rule of the form "the market costs less" does not survive them, and its opposite fails as well.
What happens to a 50/50 settlement compared with a refund?
Polymarket's resolution rules say a market that cannot be settled Yes or No resolves 50/50, and every share then redeems at 0.50. BetInAsia (help article updated 2026-09-18) and Sportmarket both describe this for cancelled events, ties and forfeits, and both say that an abandoned match goes to the player who was ahead.
A 50/50 is not a refund. A voided sportsbook bet usually returns the stake, whatever it cost you, though the rule is set book by book. A 50/50 pays 0.50 per share, so the result depends on your entry price: +0.20 per share if you bought at 0.30, zero at 0.50, -0.20 at 0.70, all before fees. BetInAsia's article makes the same point, noting that the loss comes although the match produced no result.
Delays are where the two brokers part ways. BetInAsia files some delayed matches under its Polymarket 50/50 cases, while Sportmarket says a postponement or delay alone never settles a bet, so the two do not describe delays the same way and the market's own rules are the place to look. Sportmarket also lists PMK, another prediction market, and says it settles at a fair settlement price that is not a fixed 50/50. And what would a 50/50 do to a position bought at the price you usually pay?
What stays outside this comparison?
Sportmarket says a live bet on a prediction market is taken instantly and stays valid through a goal, while bookmakers suspend; that is a broker's description of a structural difference, and this page cannot price it. The same broker shows prediction-market odds in decimal with the fee built into them, without saying how or at what rate, and warns that prices are not always directly comparable across its bet slip. If you read prices through a broker, find out which number you are copying before you feed it to the tool.
Pinnacle published its own article on the subject on 13 February 2026, framing prediction markets as trading in percentages instead of odds. Read a market's rules before buying, and check Polymarket's own access rules for your country first.
Whatever the tool shows, one habit keeps the comparison honest: check the price on two routes at the same moment, as comparing one price across two routes does for a sportsbook. When the event can be cancelled, the settlement rule may matter more than the fee.