How to Remove the Vig From Betting Odds: A 2026 Guide

To remove the vig from betting odds, convert each posted price into an implied probability, then rescale those probabilities so they sum to exactly 100%. That rescaling step is called devigging, and the method you pick (multiplicative, additive, Shin, or power) decides how the bookmaker's margin is pulled back out of each outcome. On balanced markets the methods agree almost perfectly. On lopsided markets they can disagree by more than a full probability point, which is frequently larger than the edge a bettor is trying to detect in the first place.

That disagreement is why the topic deserves more than a formula dump. Every fair-value model, every +EV screen, and every closing line value calculation has a devig step buried inside it. The step looks mechanical. The choice behind it is a modeling assumption about where the bookmaker hides its margin, and different assumptions produce different edges from identical prices.

What is the vig, and how do you measure it?

The vig (also juice, hold, margin, or overround, with some sloppiness among those terms in practice) is the bookmaker's built-in commission. It exists because the implied probabilities of a market's prices sum to more than 100%. The excess is the book's expected take.

Converting American odds to implied probability takes one of two formulas. For a negative price of -A, the implied probability is A / (A + 100). For a positive price of +B, it is 100 / (B + 100).

Apply that to the most common price in US sports betting, a point spread at -110 on both sides:

• Each side implies 110 / 210 = 52.38%.
• The two sides sum to 104.76%. That sum is the booksum.
• The overround is the excess above 100%: 4.76 percentage points.
• The hold, the book's expected share of balanced handle, is 4.76 / 104.76 = 4.55%.

Both sides of that spread cannot each have a 52.38% chance of covering. The extra 4.76 points are margin, distributed across the two outcomes in some proportion only the bookmaker knows. Devigging is the act of proposing that proportion and backing the margin out.

Why devig before measuring edge?

Raw implied probabilities overstate every outcome in the market, so any calculation built on them is biased before it starts. Expected value is the clearest case:

EV = (fairProbability * decimalOdds) - 1

Feed that formula raw implied probabilities from the same book and every bet returns exactly zero: the price cancels itself out and the hold disappears from view, which is precisely how the margin hides. Feed it raw probabilities from a sharp reference book and you systematically overcount your edge, because the reference book's margin inflates every probability above fair. The fair probability in that formula has to come from a devigged market or the output is noise.

The same dependency sits inside closing line value, the standard scorecard for betting skill. CLV compares the price you took against the price a sharp anchor book (Pinnacle, in most measurement frameworks) closed at. The comparison only means something after the anchor's closing price has been devigged into a no-vig probability. Skip the step and your CLV numbers inherit the anchor's margin structure, which varies by sport, market type, and price level.

What are the four standard devig methods?

Three families of methods cover nearly all production devigging: multiplicative (also called proportional), additive, and the margin-weighting family that includes Shin's method and the power method. They differ in a single assumption: how the margin is distributed across outcomes. Notation for a two-way market: p1 and p2 are the raw implied probabilities and B = p1 + p2 is the booksum.

Multiplicative (proportional)

Divide each implied probability by the booksum:

fair_i = p_i / B

Margin comes off each outcome in proportion to its size. The favorite gives up more absolute probability than the longshot, and both give up the same percentage of themselves. This is the default in most public tools because it is one line of arithmetic and generalizes to any number of outcomes. Its weakness is its assumption: if the book loads extra margin onto longshots, and the empirical record says books do, a proportional rescale leaves that extra shading in place and overstates the longshot's fair probability.

Additive

Subtract an equal slice of the overround from every outcome. With n outcomes:

fair_i = p_i - (B - 1) / n

Equal absolute margin per outcome means the relative haircut is far heavier on longshots. A 2-point deduction barely moves a 75% favorite and transforms a 5% outsider. The method also has a sharp pathology: any outcome whose implied probability is smaller than (B - 1) / n devigs to a negative number. That happens in real markets. A 30-team championship futures board with a 130% booksum removes a flat 1 point per team, so any team priced longer than +9900 (implied under 1%) comes out below zero.

Shin's method and the power method

Shin's method comes from a pair of Economic Journal papers (Shin 1992, 1993) that model a bookmaker setting prices while facing some fraction z of bettors trading on inside information. The book protects itself by shading longshots hardest, since informed money does the most damage there. Inverting the model recovers fair probabilities:

fair_i = (sqrt(z^2 + 4 * (1 - z) * p_i^2 / B) - z) / (2 * (1 - z))

with z solved numerically so the fair probabilities sum to 1. The recovered z is interpretable as the incidence of insider money in the market, which makes Shin's approach the only one of the four with an economic story behind its margin distribution.

The power method reaches a similar shape with less machinery. Raise every implied probability to a common exponent k and solve for the k that makes the results sum to 1:

fair_i = p_i^k,  solving k such that sum(p_i^k) = 1

Whenever there is vig, k comes out above 1, and raising a small probability to an exponent above 1 shrinks it proportionally more than a large one. Longshots surrender more margin, favorites less: the favorite-longshot correction again, through a different mechanism. Both Shin and power need a one-dimensional numeric solve (bisection or Newton's method converges in a few iterations), which is why they show up in serious pipelines more often than in spreadsheet tutorials.

A worked example: one balanced market, one lopsided

Start with the balanced case, -110 / -110. Every method takes 52.38% per side to 50.00% and a fair price of +100. When the outcomes are symmetric there is nothing for the methods to disagree about, which is why devigging looks like a solved problem in most explainers.

Now a lopsided two-way market: -300 on the favorite, +250 on the underdog. Raw implied probabilities are 75.00% and 28.57%, a booksum of 103.57%.

Three observations from that table.

First, the spread between methods on the underdog is 1.22 probability points (27.59% versus 26.37%), which translates to nearly 17 cents of American price (+262.5 versus +279.2). Bettors hunting +EV typically fire on edges of 1 to 3%. The devig choice alone spans that entire window.

Second, Shin lands almost exactly on the additive answer here. That convergence is specific to two-outcome markets. On bigger outcome sets Shin and additive separate, and additive is the one that starts producing negative numbers.

Third, the practical consequence. Suppose another book hangs +280 on this underdog. Under multiplicative fair probabilities the bet has an EV of (0.2759 * 3.80) - 1 = +4.8%. Under power fair probabilities it is (0.2637 * 3.80) - 1 = +0.2%. Same price, same market, and one method calls it a strong play while another calls it a pass. Neither answer is provably right in isolation. What is provably wrong is quoting the +4.8% without naming the method that produced it.

How do devigged prices feed closing line value?

A CLV pipeline is a devig pipeline with a clock attached. The measurement loop: record the price and timestamp of every placement, capture the sharp anchor's final pre-event price for the same market, devig that closing price into a fair probability, and score the placement against it. Aggregated over hundreds of bets, the output separates bettors who beat the market's final opinion from bettors who feed it.

Method consistency matters more here than method choice. A CLV ledger that scores March bets with multiplicative fair values and April bets with power fair values has a discontinuity running through every aggregate. Serious pipelines pick one method per market class, document it, and hold it fixed. The honest ones also publish which method they use, since the table above shows the choice moves the numbers by more than most bettors' measured edge.

This is the shape of the pipeline behind CLV.gg, the live sports-betting intelligence product we build at ixprt: real-time prices in, devigged fair probabilities per market, edges surfaced against those fair values, and every tracked bet scored for closing line value against a devigged sharp close. The engineering that keeps that loop running at production latency is covered in our walkthrough of the CLV.gg architecture.

What are the common devigging mistakes?

Devigging a stale line. The math executes identically on a fresh price and a six-hour-old one. If the anchor book moved from -300 to -350 since your snapshot, your carefully devigged fair probability describes a market that no longer exists, and every edge measured against it is fiction. For CLV specifically, the close means the anchor's last available pre-event price. For live edge detection, staleness tolerance is measured in seconds. Feed freshness is a data-engineering problem sitting upstream of the probability math, and it fails silently.

Ignoring the favorite-longshot bias. The empirical record here is long: longshots return less per dollar than favorites at equivalent implied probabilities, a pattern documented since Griffith's 1949 horse-track study and surveyed in Thaler and Ziemba (1988). Bookmakers shade longshot prices accordingly. Multiplicative devigging assumes the shading is proportional, so it overstates longshot fair probabilities, and a +EV screen built on it will keep surfacing +400-and-longer prices that are less valuable than they score. A screen that lights up disproportionately on longshots is describing its own devig method.

Using two-way math on three-way markets. Soccer 1X2 markets (home, draw, away) spread their margin across three outcomes. Devigging home and away as a complementary pair, with the shortcut fair = p1 / (p1 + p2), ignores the 25 to 30% of probability mass sitting on the draw and produces confidently wrong numbers. The two-outcome form of Shin's formula has the same restriction. Every method above generalizes correctly to n outcomes; the mistake is applying a two-outcome implementation to a market that has three. Derivative markets (draw no bet, double chance) need fair values derived from the full three-way devig, never from the derivative's own posted prices alone.

Removing the vig is a two-step operation: prices to implied probabilities, then a normalization whose method encodes an assumption about where the margin lives. Multiplicative is the transparent default. Additive, Shin, and power take progressively stronger positions on longshot shading, and their differences concentrate exactly where recreational money concentrates.

For a bettor building a fair-value habit, the order of operations is straightforward. Learn the multiplicative rescale until it is automatic. Adopt power or Shin before trusting any longshot edge. Never compare numbers produced by different methods. The formulas fit on an index card; the discipline around freshness, consistency, and outcome-set coverage is where measured edge is won or lost.