How fair value is computed

Every price on StackDubs is compared with a fair price. This page says exactly how that fair price is built, so you can decide how much to trust it.

1. Every price gets an identity

Each sportsbook line is first mapped to one canonical market: the game, the market type (moneyline, spread, total, team total, player prop), the side, the period, and the exact line. A spread of +3.5 and one of +4 are different markets. If a book's row cannot be identified with certainty, it is dropped and counted, never guessed.

2. Stale and implausible prices are removed

A pregame price older than two minutes (ten minutes when the game is more than a day away), a live price older than fifteen seconds, or an exchange ladder older than five seconds is treated as stale and masked. A two-sided market whose hold is zero or negative, or above 12 % (4 % for sharp books), is excluded. A sudden move of more than three standard deviations against its own ten-minute history is held as suspect until another book confirms it within thirty seconds.

3. The fair is a weighted, de-vigged consensus

We remove the vig from each remaining book's two sides (Shin's method when a longshot bias is visible, the power method otherwise, multiplicative as the fallback) and average the results with fixed weights: Pinnacle 1.0, Circa 0.9, BetOnline 0.7, an exchange 0.8 when at least $2,000 is resting within two cents of the best price on both sides (0.3 otherwise), and each recreational book 0.15, with all recreational books together capped at 0.3. At least two sources are required. The row shows which tier produced the fair — sharp, exchange or consensus — and how many sources were behind it.

4. Lines that differ from the fair line are translated, not forced

If a book posts a different spread or total than the one the fair was built on, its fair is translated through the sport's push probabilities (how often a game lands on 3, 7, or a given total). Translations across more than a few key numbers are refused and the row says so.

5. EV, rank and what the number means

EV is the fair probability times the book's decimal price, minus one. A consensus-tier fair must beat 4 % before a row is treated as an edge; sharp and exchange tiers 2 %. Rows are ranked by EV scaled by how much the book is believed to accept (its limit), so a 6 % edge at a book that takes $50 ranks below a 3 % edge at one that takes $2,000. Sharp reference books are never shown as places to bet.

6. We publish the receipts

Every row the scanners emit is stored and later measured against the sharp closing line. The receipts page shows, per tool and sport, how often those rows beat the close, by how much, with the sample size and the uncertainty. A tool whose rows do not beat the close is marked informational on its page and sends no alerts until it does.

StackDubs is analytics only and takes no wagers. Fair value is an estimate, not a prediction. 21+. 1-800-GAMBLER.