MargusMargus Research · Original data
Which sportsbook has the best odds?
Margus compared the last pregame moneyline at 8 sportsbooks and prediction markets on 2,093 games (September 27, 2026 to November 14, 2026): sportsbooks' median two-way moneyline margin was 4.4% and prediction markets' was 4.3% after fees, and ProphetX had the best price most often (77.0% of sides).
Sportsbook Odds Comparison 2026: 2,093 Games, 8 Books
Key findings
The median two-way moneyline at a sportsbook carried a 4.4% margin (2,616 markets); at prediction markets it was 4.3% after taker fees (1.0% on listed prices, 3,369 markets), about the same as sportsbooks.
Link to this findingOf the 8 books with at least 20 priced games, ProphetX had the lowest median moneyline margin (0.9%) and Polymarket US the highest (5.7% after fees).
Link to this findingProphetX had the best moneyline price (or tied for it) on 619 of the 804 sides it priced alongside another book (77.0%), the most of any book studied.
Link to this findingBooks disagreed on the price on 4,705 of 4,713 game sides that at least two books priced (99.8%); on the median side, the best and worst prices were 2.1 points of implied probability apart.
Link to this findingSoccer's three-way moneyline (home, draw, away) carried a median 6.7% margin at sportsbooks across 1,589 markets, against 4.4% for two-way moneylines.
Link to this findingOf 1,589 apparent moneyline, spread and total arbitrage opportunities Margus's detector flagged between October 18, 2026 and November 13, 2026, 2 (0.1%) passed every check and 1 still held when re-checked; 1,576 (99.2%) relied on at least one stale price.
Link to this finding
Moneyline margin and best-price share by book
Books with at least 20 complete pregame moneyline markets. Lower margin is better for the bettor. Best-price share counts sides the book priced alongside at least one other book, ties included.
| Book | Type | Markets | Median margin (listed) | Median margin after fees | Best price or tied | Sides compared |
|---|---|---|---|---|---|---|
| ProphetX | Prediction market | 373 | 0.9% | same as listed | 77.0% | 804 |
| Predict.fun | Prediction market | 178 | 2.6% | same as listed | 15.6% | 378 |
| DraftKings | Sportsbook | 1,155 | 4.3% | same as listed | 28.3% | 2,811 |
| Kalshi | Prediction market | 1,968 | 1.0% | 4.3% | 35.5% | 4,455 |
| BetMGM | Sportsbook | 107 | 4.4% | same as listed | 33.2% | 214 |
| theScore Bet | Sportsbook | 1,623 | 5.1% | same as listed | 28.6% | 3,763 |
| BetRivers | Sportsbook | 1,320 | 5.5% | same as listed | 19.6% | 3,160 |
| Polymarket US | Prediction market | 1,961 | 1.0% | 5.7% | 7.5% | 4,399 |
Why apparent arbitrage opportunities failed
One candidate can fail several checks, so the shares add up to more than 100%.
| Check failed | Candidates | Share |
|---|---|---|
| A price was older than the freshness limit | 1,576 | 99.2% |
| The legs were observed too far apart in time | 1,575 | 99.1% |
| Both legs were at the same book | 456 | 28.7% |
| Return too large to be real (a likely data error) | 263 | 16.6% |
| Return below the minimum after fees | 131 | 8.2% |
| A leg was a live (in-game) price | 25 | 1.6% |
| The legs had different lines | 1 | 0.1% |
| A price went stale before the re-check | 1 | 0.1% |
Methodology
- Source: prices Margus collects directly from each book's public odds feed. For every game and book, the study takes the last full-game moneyline observed before the scheduled start.
- Margin is the sum of the implied probabilities of all outcomes minus 100%: two outcomes in most sports, three (home, draw, away) in soccer.
- Kalshi and Polymarket add a taker fee to the listed price (7% × p × (1 − p) per contract). "After fees" uses the price a bettor really pays. Predict.fun quotes include its fee; ProphetX prices are used as listed.
- Excluded: 18 markets missing a side, 10 markets whose implied probabilities add up below 99.5% or above 125% (a stale or mismatched side), 0 duplicate prices, and books with fewer than 20 complete markets.
- The best price on a side is the highest payout after fees among books with a complete market for that game. Ties count for every tied book.
- Arbitrage figures come from the log of Margus's arbitrage detector, using the checks it applied at the time each candidate was flagged.
- Shares and margins are rounded to one decimal place. The study is recomputed once a day.
Suggested citation
Margus Research, "Sportsbook Odds Comparison 2026: 2,093 Games, 8 Books", https://margus.app/research/sportsbook-odds (November 2026).
Also available as markdown. Every number on this page comes from one calculation that the markdown version, the FAQ and llms.txt also use.
Frequently asked questions
Which sportsbook has the best odds?
ProphetX had the best moneyline price (or tied for it) on 619 of the 804 sides it priced alongside another book (77.0%), the most of any book studied. Of the 8 books with at least 20 priced games, ProphetX had the lowest median moneyline margin (0.9%) and Polymarket US the highest (5.7% after fees). No book is best on every game, which is why comparing prices before each bet matters more than picking one book.
How much vig do sportsbooks charge on moneylines?
In Margus's pregame prices, the median two-way moneyline at a sportsbook had a 4.4% margin: the amount the two sides' implied probabilities add up to above 100%. A standard -110/-110 market is a 4.8% margin.
Do prediction markets have better odds than sportsbooks?
The median two-way moneyline at a sportsbook carried a 4.4% margin (2,616 markets); at prediction markets it was 4.3% after taker fees (1.0% on listed prices, 3,369 markets), about the same as sportsbooks. Kalshi and Polymarket charge a taker fee on top of the listed price, so their listed prices look cheaper than what a bettor actually pays.
Are sports betting arbitrage opportunities real?
Rarely. Of 1,589 apparent moneyline, spread and total arbitrage opportunities Margus's detector flagged between October 18, 2026 and November 13, 2026, 2 (0.1%) passed every check and 1 still held when re-checked; 1,576 (99.2%) relied on at least one stale price. Most apparent arbs come from one book's price being out of date.
More original data