No-Vig Odds: How to Calculate Them—and Why They’re Estimates

No-vig odds are estimated fair prices after removing a sportsbook’s margin from a complete market. Using proportional margin removal, illustrative decimal odds of 2.00, 3.00 and 4.00 imply probabilities totaling 108.33%; normalizing them produces 46.15%, 30.77% and 23.08%, but those estimates are not guaranteed to reflect true chances.
The calculation is straightforward; interpreting it takes more care. If you need the underlying terminology first, read our guide to vig and the bookmaker’s margin. Here, the focus is turning a complete set of prices into a usable benchmark—and understanding what that benchmark cannot tell you.
A quoted betting price implies a probability. For decimal odds, implied probability equals one divided by the odds. When the implied probabilities of every mutually exclusive, collectively exhaustive outcome add up to more than 100%, the excess is the market’s overround.
No-vig probabilities remove that excess according to a chosen rule. The proportional, or multiplicative, method divides each implied probability by the total. No-vig decimal odds are then the reciprocals of those normalized probabilities, as explained by the StatsNext no-vig calculator.
“Fair odds” therefore means fair under the conversion method and its assumptions. It does not mean that an independent model has established the outcome’s true probability, or that a bet at a better price must win.
Use every outcome from the same sportsbook, market, line and observation time. For a three-way football result market, that means home win, draw and away win under identical settlement rules. The following prices are an illustrative calculation, not live odds or a historical Pinnacle quote.
Keep full precision during the calculation and round only the displayed results. You can use the odds to probability calculator to check the initial conversions. These raw probabilities include the margin; they are not yet a coherent probability distribution because they exceed 100% together.
The total implied probability is 0.50 + 0.333333… + 0.25 = 1.083333…, or 108.33%. The overround is therefore 108.33% − 100% = 8.33 percentage points. Use 1.083333… as the divisor when normalizing probabilities expressed as decimals.
The proportional no-vig formula is qᵢ = (1 ÷ dᵢ) ÷ Σ(1 ÷ dⱼ), where dᵢ is an outcome’s decimal odds and the sum includes every outcome in that market. This preserves the relative ratios between the original implied probabilities.

The normalized probabilities now total 100%. Notice that every no-vig decimal price is longer than the original quote: removing a positive proportional margin increases the hypothetical payout corresponding to each outcome.

No. StatsNext distinguishes overround, T − 1, from theoretical hold under a specified staking pattern, 1 − 1 ÷ T. In this example, the overround is 8.33%, while that hold calculation gives 7.69%.
The distinction matters because an excess probability total is not automatically the percentage of actual stakes a sportsbook will retain. The hold expression assumes action distributed to match the prices; it is not a forecast of realized revenue from an unknown collection of bets.
Removing margin solves an accounting problem: the probabilities must sum to 100%. It does not independently solve the forecasting problem of how likely each outcome really is. Many probability distributions could satisfy that total, and the normalization rule determines which one you obtain.
The multiplicative method assumes margin is distributed in proportion to each outcome’s implied probability. StatsNext explicitly describes that as a convention, not a fact about how a sportsbook sets its prices. If the underlying price distortions are not proportional, normalization will not necessarily remove them.
In the example, 46.15% is therefore the home-win estimate produced by this method—not a measured property of the match. Reporting more decimal places would make the arithmetic more precise without making the forecast more certain.
Yes. The paper Forecast Sports Outcomes under Efficient Market Hypothesis identifies Multiplicative, Shin and Power as existing odds-only conversion methods. It describes converting betting odds into accurate probabilities as a fundamental challenge, rather than a settled consequence of removing overround.
The paper proposes another odds-only method without historical data for model fitting and reports that it outperformed existing odds-only methods for most bookmakers in its dataset. That is a dataset-specific finding, not proof that one method always wins. The practical lesson is that the conversion method can affect a marginal value decision.
The OddsFantasy positive EV finder lists pre-match value bets from sportsbooks, betting exchanges and prediction markets, priced against Pinnacle’s no-vig odds. Each listing includes its expected value and a link to the bookmaker. This makes the benchmark actionable without turning it into certainty.
To illustrate the comparison, suppose a benchmark assigns the home win the same 46.15% probability calculated above, while another sportsbook offers 2.30. This is a hypothetical comparison, not a claim about a current listing or the specific margin-removal implementation used by OddsFantasy.
Expected net return per unit staked is p × d − 1. Using the unrounded probability, (6 ÷ 13) × 2.30 − 1 = 0.061538…, or approximately +6.15% estimated EV before any applicable costs. The break-even probability at 2.30 is 43.48%, below the benchmark’s 46.15%.
You can reproduce that calculation with the expected value calculator. If the real probability were below 43.48%, however, the same bet would have negative EV. A favorable comparison is conditional on the quality of the probability estimate and the price you actually obtain.
OddsFantasy finds and alerts sportsbook value bets; users place those bets themselves. Automatic value-bet placement is available only on Polymarket and requires an automation key saved in the user’s trading wallet.
Polymarket automation can filter by minimum and maximum EV, odds range, sports, maximum spread, minimum volume and stake per market. It uses a resting limit order by default. With market entry enabled, it fills immediately and skips the bet if the price has moved beyond the slippage limit. These controls manage execution conditions; they do not validate the benchmark probability.
For parlays, no-vig probabilities for individual legs also do not establish their joint probability; dependence still matters. Across singles, accumulators and prediction-market trades, use the no-vig line as a disciplined reference price—not a guarantee of accuracy or profit.
Convert every outcome’s decimal odds to implied probability using 1 ÷ odds, add those probabilities, and divide each by the total. Take the reciprocal of each normalized probability to obtain no-vig decimal odds. Use all outcomes from the same sportsbook, market and line.
For decimal odds, implied probability = 1 ÷ decimal odds. For example, 4.00 implies 25%. This is the raw price-implied probability; removing margin requires the prices of every outcome in the market.
Odds specify the payout available on a bet, while probability expresses an outcome’s estimated likelihood. Decimal odds of 3.00 mean a winning one-unit bet returns three units including the stake. Their reciprocal, 33.33%, is the break-even probability before costs—not necessarily the true chance.
Not using complete-market proportional normalization. A single price gives its implied probability, but you need every outcome’s price to calculate the market’s total implied probability and remove its proportional margin.
No. A better price can indicate positive expected value relative to that benchmark, but the benchmark is an estimate. Probability error, changed prices, execution costs and losing outcomes can all affect results.

Compare sharp no-vig probabilities with executable Polymarket prices, then account for fees, liquidity and stale odds before autom...

What the vig (vigorish or juice) is, why bookmakers build it into every set of odds and how it affects the price you get, with rea...

How to read decimal, fractional and American betting odds, what each format says about probability and payout, and how to interpre...