No-Vig Calculator: Remove Bookmaker Margin & Calculate Fair Odds

No-Vig Calculator & Fair Odds Calculator

Enter a complete market from one bookmaker and one snapshot. The tool converts each price to raw implied probability, calculates booksum, overround and margin as a share of booksum, then returns no-vig probabilities and fair decimal odds. Proportional is the transparent baseline; Shin is the main sensitivity comparison, while power, additive and odds-ratio remain available for method comparison.

Do not mix overlapping bets or best prices from different bookmakers. If a correct-score or outright market defines a catch-all outcome such as “Any Other Result,” include it.

From decimal odds to raw implied probabilities

Need to convert one price rather than remove margin from a complete market? Use the Implied Probability Calculator for decimal, American or fractional odds, or to convert probability back to equivalent odds.

For decimal odds O_i:

For a three-way football market quoted at 2.10, 3.50 and 3.40:

The overround is:

Raw reciprocals are not a proper probability distribution because they do not sum to one. They are the starting point for every method below.

Why the margin-free probabilities are not uniquely identified

A three-outcome market provides three quoted odds. The unknowns are the three fair probabilities plus a rule describing how the bookmaker transformed those probabilities into prices.

The condition:

is necessary, but it does not specify how the 5.602-point excess should be allocated. Several 100% distributions are compatible with the same odds under different pricing assumptions.

Margin-removal methods at a glance

1. Proportional or multiplicative normalization

Divide each raw probability by the booksum:

In the example, Q = 1.056022:

What it assumes

The same proportion of each raw implied probability is attributed to margin. Every fair probability equals approximately 94.7% of its raw value in this example.

Strengths

• closed-form calculation;
• always sums to 100%;
• preserves the ratios between raw probabilities;
• easy to explain and reproduce;
• appropriate as a neutral baseline.

Limitation

If bookmakers apply higher relative margins to longshots than favourites, proportional normalization can leave favourites too low and outsiders too high.

2. Additive method

Subtract an equal share of the overround from every outcome:

With three outcomes and 5.602 points of overround, each outcome loses:

Effect on favourites and longshots

Subtracting 1.867 points is a small relative change for a 47.6% favourite but a larger relative change for a 28.6% outcome. The method therefore shifts probability toward the favourite relative to proportional normalization.

Failure condition

Additive probabilities become negative when:

This can occur in outrights, correct-score markets and other books containing very long-priced outcomes.

3. Power method

Raise every raw implied probability to the same exponent α:

Choose α so that:

For a positive overround, α is normally greater than 1. Since all q values lie below 1, exponentiation reduces them. Smaller longshot probabilities generally shrink more in relative terms than larger favourite probabilities.

Example parameter

For 2.10 / 3.50 / 3.40:

Strengths

• probabilities remain within 0 and 1;
• naturally allows non-uniform relative adjustment;
• only one scalar parameter;
• simple bisection gives a stable solution.

Limitation

The power shape is convenient, but one exponent is still only an assumption about how margin relates to event probability.

4. Odds-ratio method

Convert raw probability q to event odds q/(1−q), apply a common divisor c, and convert back to probability:

Choose c so the probabilities sum to one.

Logit interpretation

The method applies the same shift to every outcome on the log-odds scale rather than the probability scale.

Example parameter

For the example market:

Interpretation

Compared with proportional normalization, the odds-ratio method typically raises the favourite and reduces longshots, but often less aggressively than additive or power for ordinary three-way football markets.

5. Shin method

Shin’s framework models a bookmaker facing a mixture of ordinary bettors and an assumed informed component. The commonly used event-level inversion calculates:

The parameter z is chosen so:

Example parameter

For 2.10 / 3.50 / 3.40:

The resulting distribution lies between the proportional and the more favourite-weighted additive/power estimates in this example.

Do not interpret z literally without qualification

The original model gives z an informed-bettor interpretation. More recent theoretical work argues that event-level z estimates can be positive even when there are no insiders and can reflect overround, bookmaker costs and market competition instead.

Worked example: five methods on the same football odds

Use the market 2.10 / 3.50 / 3.40.

How sensitive are the results to the chosen method?

This range measures model sensitivity only — how much the answer moves when the de-vig method changes. It is not a confidence interval for the true probability.

The differences are modest in this balanced, moderate-margin market. They can become larger with a stronger favourite, more outcomes or a higher overround.

Why the favourite increases under four methods

Compared with proportional normalization, additive, power, odds-ratio and Shin all move some probability from the two longer-priced outcomes toward the home favourite in this example. They differ in the exact shape and theoretical justification of that movement.

What changes in a two-outcome market?

Examples include BTTS Yes/No, Over/Under 2.5 and many team-total markets.

Shin equals additive for two outcomes

For an ordinary two-outcome book, the event-level Shin probabilities reduce to:

Therefore, showing separate Shin and additive columns in a binary de-vig calculator can create a false impression of independent methods when they are mathematically identical under this formulation.

Power and odds-ratio remain distinct

Power and odds-ratio transformations can still differ from additive in a lopsided binary market.

Which bookmaker margin-removal method should you use?

Use proportional when:

• transparency and reproducibility are the main requirements;
• the result is a baseline rather than a claim of optimal forecasting;
• the market is balanced and low-margin;
• you need a stable method across many market types.

Use power when:

• you want a one-parameter favourite–longshot adjustment;
• additive can create negative probabilities;
• you need outputs guaranteed inside [0,1] for ordinary positive-overround books;
• out-of-sample tests support it for the bookmaker and market.

Use Shin when:

• you need continuity with research using Shin probabilities;
• you are analysing three-or-more-outcome markets;
• the method has been validated on the relevant data;
• z is reported as a model parameter rather than a literal insider share.

Use additive when:

• the market has few outcomes and no probability approaches zero;
• you explicitly want equal probability-point removal;
• you check for negative outputs.

Use odds-ratio when:

• a constant logit-scale adjustment is substantively reasonable;
• you want an alternative smooth transformation;
• you can validate it against the other methods.

How to test which method is better

1. Freeze odds before the event. Use a consistent source and timestamp such as closing odds.
2. Run every method on the same books. Do not compare methods on different match samples.
3. Store full probability vectors. For football 1X2, retain Home, Draw and Away for every match.
4. Use proper scoring rules. Compare multiclass Brier score, log loss or ranked probability score where justified.
5. Check calibration separately. Build reliability diagnostics for favourites, draws and away outcomes.
6. Test favourite–longshot residuals. Determine whether low-probability outcomes remain systematically too high or too low.
7. Use paired uncertainty. Bootstrap or otherwise evaluate match-level score differences.
8. Repeat by bookmaker and market. A method that works on one source may not transfer.

What existing studies show

Štrumbelj’s multi-sport comparison found Shin probabilities more accurate than basic normalization in the analysed bookmaker/competition pairs. A separate comparison by Clarke, Kovalchik and Ingram found the power method outperformed proportional normalization across their datasets and was generally comparable with or better than Shin. These findings are evidence for those samples, not a universal ranking.

More recent football research continues to propose alternatives that outperform older odds-only transformations on specific large datasets. This reinforces the need to validate the method rather than select one by reputation.

See football probability calibration for Brier score, log loss and reliability diagrams.

What this means for ImpliedScore

ImpliedScore currently uses proportional normalization because it is transparent, deterministic and easy for users to reproduce. The normalized 1X2 and Over/Under probabilities become targets for fitting λ_home and λ_away.

Why changing de-vig method changes the score matrix

A more favourite-weighted Home/Draw/Away distribution can increase the gap between the two fitted goal rates. A different Over/Under probability can change their sum. Therefore:

Method choice can affect:

• home and away expected goals;
• 1X2 residuals;
• most likely score;
• BTTS and team totals;
• fair odds for derived markets.

Production baseline versus comparison methods

The public calculator above can compare several de-vig transformations, but that does not mean ImpliedScore silently switches its production score engine or historical tracker data between them. Proportional normalization remains the reproducible production baseline. Shin, power, additive and odds-ratio are sensitivity/comparison outputs until paired out-of-sample validation justifies a different production choice.

1. Keep proportional as the visible production baseline.
2. Use Shin as the primary non-proportional sensitivity comparison.
3. Label every alternative output with its method.
4. Do not rewrite historical tracker snapshots when a different method is selected in this calculator.
5. Store an explicit method identifier if a future tracker/export uses a non-proportional method.
6. Validate any production-method change on frozen odds and final outcomes before deployment.

The next modelling stage is explained in how to infer expected goals from betting odds.

Practical conclusion

References

1. Štrumbelj, E. (2014). On Determining Probability Forecasts from Betting Odds . International Journal of Forecasting, 30(4), 934–943.
2. Clarke, S. R., Kovalchik, S. & Ingram, M. (2017). Adjusting Bookmaker’s Odds to Allow for Overround . American Journal of Sports Science, 5(6), 45–49.
3. Shin, H. S. (1992). Prices of State Contingent Claims with Insider Traders, and the Favourite-Longshot Bias . The Economic Journal, 102(411), 426–435.
4. Shin, H. S. (1993). Measuring the Incidence of Insider Trading in a Market for State-Contingent Claims . The Economic Journal, 103(420), 1141–1153.
5. Whelan, K. (2024). On Estimates of Insider Trading in Sports Betting . University College Dublin working paper.
6. Whelan, K. & Hegarty, T. (2023). Calculating the Bookmaker’s Margin: Why Bets Lose More on Average Than You Are Warned . MPRA Paper 116924.
7. Goto, K., Takeishi, N. & Yairi, T. (2026). Forecast Sports Outcomes under Efficient Market Hypothesis: Theoretical and Experimental Analysis of Odds-Only and Generalised Linear Models . arXiv preprint arXiv:2604.17194.
8. Cheung, K. (2015). Fixed-Odds Betting and Traditional Odds . Sports Trading Network. Source of the equal-odds-ratio margin distribution and removal method.
9. ImpliedScore. Methodology: From Odds to a Score Probability Matrix.