Closing Line Value by Book

We tracked every +EV line our scanner flagged across every sportsbook and exchange we cover, then recorded where the market closed. This is which books' mispricings held to the close and which ones got corrected before the whistle.

No sportsbook pays Iron Marker a referral fee. There is no book we are being paid to keep off the bottom of this table.

Closing line value by book

Closing line value is the price you took, valued against where the market ended up. For every flagged bet we compare the price on the board at the moment the scanner surfaced it against Pinnacle's no-vig fair probability at the last pre-game reading. A positive number means the flagged price was better than the close. It is the standard test of whether an edge was real, because the closing line is the sharpest price a market produces.

Read the median column. Means run above it because a small number of thin-market moneylines carry very large CLV: BetAnySports is the clearest case, a mean of +19.70% against a median of +2.62% on 64 bets, a gap of 17.08 points. Trimming 5% from each tail brings that mean to +6.79%. The mean and trimmed-mean columns are published so the tail is visible rather than hidden.

Four books cleared closing-line capture but not the 50-bet floor for this table, so they appear in the CSV only, flagged below_min_n: BetMGM (n=47), ReBet (n=40), Fanatics (n=31), Fliff (n=7).

By market

Splitting the same bets by market type separates where the edge is large from where it is dependable.

Moneyline edges are the largest and the least dependable: a median +1.70% on 2,933 bets, but a mean of +5.79% that a 5% trim cuts to +2.21%. Spreads and totals are smaller and steadier: totals held a median +1.32% on 2,067 bets with 70.4% beating the close, the highest beat rate of the three, and spreads a median +1.26% on 1,920 bets at 65.5%. The market with the biggest headline number is also the one where any single result tells you the least.

By segment

Books fall into three groups that behave differently: state-licensed US sportsbooks, peer-to-peer exchanges and prediction markets, and offshore books. The exact membership of each group is listed in the methodology below.

Mispricings found at regulated US sportsbooks held up best: a median +2.18% across 1,288 flagged bets, with 70.3% beating the close. Exchanges and prediction markets came in at +1.29% (n=4,795) and offshore sportsbooks at +1.26% (n=837).

Volume runs the other way. Regulated US sportsbooks produced 18.6% of the flagged bets in this sample, exchanges and prediction markets 69.3%, and offshore sportsbooks 12.1%. That is a statement about where this scanner found flaggable prices in this window, not a ranking of how well each segment prices its whole board.

Methodology

• Data. Every row Iron Marker's scanner wrote to its flagged-bet ledger that carries a closing-line snapshot: 7,807 rows at the pull, of which 6,920 were used. Closing-line capture shipped on August 27, 2026, so the earliest event in the sample commenced August 27, 2026 and the latest September 8, 2026. Pulled 2026-09-08 13:20 UTC. Across 23 books and exchanges, 19 cleared the 50-bet floor for the table above.
• What CLV means here. For each flagged bet, clv_pct = (price at first sighting × Pinnacle no-vig fair probability at the close − 1) × 100. The price is the one posted at that book when the scanner surfaced the edge. The fair probability is Pinnacle's two-way market for the same event, side and number, devigged multiplicatively, read from the last snapshot taken before the event started. See closing line value and no-vig fair odds for the arithmetic.
• What CLV does not mean here. This is not a book's own closing price. Our ledger's field for the flagged book's price at the close is empty on every row, a known gap in the capture, so nothing on this page can say where an individual book closed a number or by how much it moved. Every figure compares a first-sighting price to the Pinnacle fair close and nothing else.
• Selection. These are only lines the scanner flagged as +EV at that book. The sample is not a book's board, it is the subset of that board we thought was mispriced. Read every figure as "how much of a flagged edge survived to close at this book", never as "how sharp this book is".
• Window and inclusion. 887 rows carried a CLV number but had not commenced at the pull and were excluded. The closing snapshot is overwritten on every scan tick until an event starts, so it is a final close only once the event is under way. Bet outcome is irrelevant to CLV, so won, lost, push, unscored and still-pending rows all count toward it: of the 6,920 used, 2,629 won, 3,032 lost, 62 pushed, 590 went unscored and 607 were still awaiting settlement.
• Median, mean and trimmed mean. Median is the headline because CLV distributions are skewed: thin moneyline markets in combat sports throw very large positive readings that pull a mean far above anything a bettor would experience. The mean is published alongside so the skew is visible, and the trimmed mean discards the top and bottom 5% of each sample.
• Beat the close. The share of a sample with CLV strictly above zero. Exactly zero counts as not beating it.
• Hit rate. Won divided by won plus lost, ignoring pushes and unsettled bets. Printed only where at least 100 bets have settled; below that it reads n/a. Across the whole sample the hit rate was 46.4% on 5,661 settled bets. A hit rate under 50% is normal and is not evidence against the edge, because most flagged bets are priced above even money; CLV, not win rate, is the measure here.
• Table floor. A book needs at least 50 flagged bets in the window to appear in the table above. Smaller books are in the CSV with a below_min_n flag rather than on the page, because a median over a few dozen bets moves too much to publish next to one over a thousand.
• Segments. Assigned per book, in the generator, as follows:
  • Regulated US sportsbooks: ballybet, betmgm, betparx, betrivers, draftkings, espnbet, fanatics, fanduel, fliff, hardrockbet, williamhill_us
  • Exchanges and prediction markets: betopenly, kalshi, novig, polymarket, prophetx, rebet
  • Offshore sportsbooks: betanysports, betonlineag, betus, bovada, lowvig, mybookieag
  State-specific variants of the same operator are collapsed onto the parent book. Pinnacle is the fair-line baseline and is never scored as a book here.
• Reproducibility. Every number on this page and in the CSV is written by tools/research/closing-line-by-book.mjs from the ledger, with the statistics in worker/lib/clvByBook.js under unit test. Nothing here is typed by hand or estimated.

What to do about it

The practical reading is not "bet at the book at the top of the table". It is narrower and more useful than that.

• Track CLV, not last week's profit. Two thirds of these flagged bets beat the close while the settled hit rate sits under 50%. That combination is exactly what a real edge looks like on a short sample: the wins have not arrived yet, but the prices were right. If your own bets are not beating the close, the results will not save you.
• Take the number fast. The edges that survive are the ones taken at first sighting. That is what the price in this dataset is, and the gap between the flagged price and the close is what disappears while you deliberate.
• Do not treat all markets alike. Moneyline edges are the biggest and the most outlier-driven. Spread and total edges are smaller and steadier. Size accordingly, with a Kelly stake rather than a flat one.
• Have accounts in more than one segment. Regulated books, exchanges and offshore books did not produce the same edges in the same weeks. A bettor limited to one app sees whichever slice that app happened to misprice.

If any of the terms above are new, start with what +EV betting is, then closing line value and line shopping.

Download the data

The CSV carries every cut on this page plus the books below the table floor and a per-sport breakdown. Columns: scope, key, label, segment, n, median CLV, mean CLV, 5% trimmed mean, beat-the-close rate, won, lost, push, settled, hit rate, and the small-sample flag. Free to use with attribution to Iron Marker (ironmarker.com). For questions about the data or custom cuts, email [email protected].

Download CSV

More research

This page measures what happens after you find the edge. Our other study measures what it costs not to look for one: The Cost of Loyalty Index puts a dollar figure, per $1,000 wagered, on betting at a single sportsbook instead of shopping the line.

Iron Marker is an independent analytics tool, not a sportsbook, and is not affiliated with any operator named above. Figures describe prices observed during the sampling window and will drift as books re-price and as the sample grows. Past closing line value does not predict future results. This page is educational and not betting or financial advice. Must be 21+ (age varies by state). Problem gambling? Call 1-800-GAMBLER.