Every market on the board is off the sharpest available price. That strips the and gives a no-vig fair probability with the book's built-in margin removed. If a prices the same outcome 2.5% or more above that fair number, it becomes a alert. Click any underlined term to open its definition in the glossary below.
The method
Vig (vigorish, juice, margin) is the commission a sportsbook bakes into its odds to guarantee itself a profit over time. No-vig odds represent the true statistical probability of an outcome with that commission stripped back out.
We start from the sharp books — high-volume, high-accuracy shops like Pinnacle that effectively set the true market number.
Their built-in commission is removed (de-vigged) from both sides of the market to recover the true probability and true price.
That fair price is compared against the soft books you actually bet — DraftKings, FanDuel, BetMGM and friends.
When a soft book pays 2.5%+ above fair, it's flagged as a positive expected value bet and pushed to your feed.
Quick example — a coin flip
True odds (no-vig)
+100 / +100
50% chance either way.
Book odds (with vig)
-110 / -110
The book takes a ~4.7% cut.
Soft book slips to
+105
>2.5% above fair → alert.
De-vigging calculates back to that true +100 baseline. If a soft book offers +105 on the same flip, you now hold a measurable statistical edge over the bookmaker — and that is precisely the condition EdgeDesk hunts for, all day, across every market.
Worked example
Sample numbers on a two-way NFL moneyline, run through exactly the pipeline the engine uses on every pick.
Pinnacle prices the game -108 on our side and -102 on the other. Those imply 51.92% and 50.50%.
The two probabilities sum to 102.42%, not 100% — the extra 2.42% is the book’s hold. Normalise our side by the total:
51.92% ÷ 102.42% = 50.70% fair probability
A true 50.70% chance is worth -103. Anything better than that number is a bet with positive expectation.
| Book | Price | Break-even | Edge |
|---|---|---|---|
| BetMGM | +100 | 50.00% | +1.39% |
| FanDuel | +102 | 49.50% | +2.41% |
| DraftKings | +105 | 48.78% | +3.93% |
Each price needs a different break-even win rate. Our fair probability beats all three, but only one clears the threshold by enough to be worth the variance.
DraftKings at +105 returns +3.93%, above the default 2.5% floor — so the alert fires. Quarter-Kelly sizing puts 0.94% of bankroll on it, or $18.71 on a $2,000 roll.
From there the bet is logged, and every later sync compares the price we took to the closing number to record its CLV.
Try it yourself
Enter both sides of a sharp book’s two-way market to strip its commission and recover the fair probability. Add the price your own book is offering to see the edge — the same calculation that generates every alert on the board.
Inputs
e.g. Pinnacle price on the outcome you want.
The opposite outcome at the same book.
Optional — leave blank to see fair value only.
The share link carries the exact inputs, so anyone who opens it sees this same calculation reproduced from scratch.
Results
Positive but inside the noise band; the default engine would skip it.
A two-sided market posted at -110 / -110 52.4% on each side, which sums to 104.8%. That extra 4.8% is the — the book's margin. Treating the raw implied number as a probability means you are betting into a price that is guaranteed to be too short on both sides at once, so the first step in any honest calculation is removing that margin.
Convert both American prices to implied probabilities, add them to get the overround, then divide each one by that total. A -140 / +120 market 58.3% and 45.5%, summing to 103.8%; dividing gives a fair 56.2% and 43.8%. Converting 56.2% back to American odds yields a fair line of about -128. That -128 is the number every price is measured against.
is only meaningful against a market that is actually efficient. Desk uses low-margin, high-limit books as the reference and treats promotional or recreational books as the price you are attacking, never as the source of truth. If you a against another soft book, you mostly measure the difference between two marketing departments.
With a p and a decimal price d at the , per unit staked is p x (d - 1) - (1 - p). At a fair 56.2% and a soft price of +110 (2.10 decimal), that is 0.562 x 1.10 - 0.438 = +0.18 units per unit risked, a 18% theoretical . Real alerts sit far lower: the board thres is 2.5%, and anything above roughly 12% is treated as stale data rather than free money.
Multiplicative assumes the book spreads its margin evenly across both sides. On heavy favourites and long-odds props the margin is skewed toward the longshot, so favourite-side fair probabilities come out slightly too high. That is one reason the board refuses prices longer than +398 and caps single-ticket — those are exactly the spots where the math is least trustworthy.
Before you look for s, look at . A market held at 2% is worth attacking; a same-game prop held at 12% almost never produces a real edge because the fair number sits deep inside the spread. Screening by hold first removes most of the false positives that make naive screens look far more productive than they are.