What Sportsbook or Exchange Is the Sharpest? Two Years of CLV Data

How to tell if a sportsbook or exchange is sharp

A book earns trust in two ways. Its closing price has to match what actually happens, which is accuracy, and the rest of the market has to move toward it when they disagree, which is speed. A book that is accurate and gets followed after it moves is a sharp data point, and following it becomes a pattern rather than a guess.

Every closing price in this study had the vig removed first, so each book's number is a fair percent chance rather than a price with margin built in. The no vig calculator does the same thing for a single two way market.

The closing line is accurate at every major book

At DraftKings, sides that closed around 67 percent won 68 percent of the time, and sides that closed around 89 percent won 90 percent of the time. The strongest proof came from line movement: in the 9,907 games where the line moved 3 points or more, the side it moved toward was priced at 51.2 percent at the open and 57.5 percent at the close, and it won 58.0 percent of the time.

That is why beating the closing number matters so much, as covered in why positive CLV means you're making sharp bets. The catch is that every major book passed this test, and FanDuel and DraftKings closed 12,883 pro games with nearly identical accuracy. Accuracy alone could not rank the books, so speed became the deciding test.

Who the market follows on game lines

Every line move on 6,087 games was pulled and timed to the second. For each game, each book's price 24 hours, 6 hours and 1 hour before the start was compared with everyone else's, and when a book disagreed with the rest, the study measured how far the rest of the market moved toward it by the close. At 100 percent the market always ends up where that book already was, and at 0 percent the market never moves toward it.

Novig came in at 68 percent, and six hours before a game the market closed 89 percent of its gap toward Novig. Kalshi was at 36 percent and FanDuel at 31 percent, while bet365, BetRivers and Hard Rock sat at zero or below, which means the rest of the market almost never followed them.

The sharpest books for player props

On props, the books were put head to head. When two books close the same player at the same number with different prices, the box score decides which one was closer. That comparison was graded every time two books were 2 points or more apart, which happened more than 400,000 times.

The result reads as how much of the edge was real when a price beat each book's closing number. Novig came in at 121 percent across 7,468 matchups, which means bets that beat its close won even more often than Novig itself expected. Kalshi was at 91 percent, FanDuel at 68 and DraftKings at 63, while bet365, BetRivers and Hard Rock all finished under 50 percent, so whenever they disagreed with another book, the other book had the better number.

The over tax on player props

The biggest surprise was hiding inside the player props. Across 215,532 DraftKings props, overs were priced to hit 49.6 percent after removing the vig, and they hit 48.1 percent. The same lean showed up at every major sportsbook graded.

That changes what CLV is worth. When a bet beat a sportsbook's closing price on an under, 95 percent of that edge showed up in the results. When it beat the close on an over, about a third of the edge showed up. Neither figure is a win rate: it is the share of the expected edge that turned into real wins. Kalshi showed none of that lean and Novig about half of it, which is a big reason the exchanges carry the weight on props. More on the over and under split is in unders vs overs on player props.

The sharpest books, ranked

How the results set fair value weights

Every EV number on Upside is measured against a fair price, and that fair price is a weighted blend of the books. The last step of the study was a formula that finds the blend whose fair price lands closest to the close, where a book earns weight for being early and right and earns nothing for repeating what another book already says.

On game lines that put Novig at 7.5, with FanDuel at 1.1, Kalshi at 1.1 and DraftKings at 1. On player props Novig and Kalshi both went to 7, with DraftKings at 1.4 and FanDuel at 1. Books that limit winning bettors fast, like Hard Rock, BetMGM, BetRivers and Fanatics, sit at 0.1 or are turned off no matter how their lines graded, because a price that sharp money is barely allowed to bet has never really been tested. ProphetX, Pinnacle, Polymarket, Circa and the other books with about a week of history are parked near FanDuel and DraftKings until there is enough data to rank them.

How to use these exact weights

Both sets are saved as presets, so nothing has to be typed in. Open the +EV Optimizer or the +EV Sniper, click Filters, and at the top next to Fair value click Weights. Under Recommended, pick Matt's Game Line Weights for sides and totals or Matt's Prop Weights for player props, then click Apply and Apply filters. Every EV number on the board is then measured against that weighted fair price.

Three settings help make sure the edge on the board is real. Set Required book to Novig, or Kalshi on props, so the book carrying the weight is actually pricing that line. Raise Min sportsbooks to 3 or more, so one stale price cannot create an edge by itself. And be careful with prop overs at a sportsbook, because unless Novig or Kalshi is on that line, the over tax is baked into the fair price.

Frequently asked questions