De-vigging is the process of removing the bookmaker's embedded margin from betting odds to produce a normalized estimate of the probabilities represented by the market. It is one of the fundamental techniques in quantitative sports-betting analysis because raw bookmaker probabilities normally contain an overround and therefore do not add up to exactly 100%.
Understanding de-vigging allows you to move from simply reading bookmaker prices to estimating the probability distribution that those prices imply after the embedded margin has been accounted for.
When decimal odds are converted into implied probabilities, the probabilities of all possible outcomes in a bookmaker's market normally exceed 100%.
This excess is the bookmaker's overround, also known as the margin, vig or juice.
De-vigging attempts to remove this excess and normalize the probabilities so that they add up to 100%.
For example, consider a hypothetical football 1X2 market:
| Outcome | Odds | Implied Probability |
|---|---|---|
| Home Win | 2.00 | 50.00% |
| Draw | 3.40 | 29.41% |
| Away Win | 4.00 | 25.00% |
| Total | 104.41% |
The implied probabilities total 104.41%.
Therefore:
Overround = 104.41% − 100% = 4.41%
The purpose of de-vigging is to remove that 4.41% excess and estimate the normalized probability distribution.
Raw implied probabilities can be misleading because they include the bookmaker's margin.
For example, decimal odds of 2.00 correspond to a raw implied probability of:
1 ÷ 2.00 = 50%
However, if the entire market has an overround, that 50% should not automatically be interpreted as a fair 50% probability.
The market may contain a pricing margin distributed across its possible outcomes.
De-vigging provides a way to normalize the market before comparing it with an independent probability model.
The simplest approach is proportional normalization.
The formula is:
De-Vigged Probability = Implied Probability ÷ Total Implied Probability
In the example above, the total implied probability is:
1.0441
Therefore:
Home = 0.5000 ÷ 1.0441 = 47.89%
Draw = 0.2941 ÷ 1.0441 = 28.17%
Away = 0.2500 ÷ 1.0441 = 23.94%
The normalized probabilities now add up to approximately 100%.
Once the probabilities have been normalized, they can be converted into theoretical fair odds.
The formula is:
Fair Odds = 1 ÷ De-Vigged Probability
| Outcome | De-Vigged Probability | Fair Odds |
|---|---|---|
| Home Win | 47.89% | 2.09 |
| Draw | 28.17% | 3.55 |
| Away Win | 23.94% | 4.18 |
The resulting prices represent the market's estimated fair prices under the proportional de-vigging assumption.
This distinction is essential.
De-vigging does not reveal the objectively true probability of an event.
It produces an estimate of the market's probability after removing the bookmaker's margin according to a particular methodology.
The resulting estimate can still be affected by:
Therefore, "fair odds" in this context should be understood as margin-adjusted market odds, not guaranteed true odds.
Once fair odds have been estimated, they can be compared with an available bookmaker price.
Suppose the market produces a de-vigged probability of:
47.89%
The corresponding fair odds are approximately:
2.09
If another bookmaker offers:
2.20
the available price is higher than the estimated fair price.
If the bookmaker instead offers:
1.90
the available price is lower than the estimated fair price.
This creates the fundamental comparison:
Estimated Fair Odds vs Available Market Odds
However, a difference between the two does not automatically prove that the available price represents value. The fair probability itself is an estimate and may be wrong.
De-vigging becomes particularly useful when combined with expected-value analysis.
The expected-value formula is:
EV = (Probability × Odds) − 1
Suppose the de-vigged probability is 47.89% and the available price is 2.20.
EV = (0.4789 × 2.20) − 1
EV ≈ +5.36%
Under the assumptions of the de-vigging method, the price therefore appears higher than the estimated fair price.
But positive calculated EV does not guarantee a winning outcome. It only indicates that the price is favourable relative to the probability estimate being used.
Proportional normalization is straightforward, but it is not the only method available.
Different de-vigging techniques make different assumptions about how the bookmaker's margin is distributed across outcomes.
The proportional method divides every implied probability by the total implied probability.
Fair Probability = Raw Probability ÷ Total Probability
It is simple, transparent and useful as a baseline method.
The Shin method attempts to account for the possibility that bookmakers face informed and uninformed bettors. As a result, it can distribute the estimated margin differently across outcomes rather than removing it proportionally.
The power method applies a mathematical transformation to the raw implied probabilities and estimates a scaling parameter that brings the probabilities back to 100%.
Because these methods use different assumptions, they can produce different fair-probability estimates from the same market.
Consider a market with a strong favourite and several much larger-priced outcomes.
A proportional method assumes that the bookmaker's margin can be removed proportionally. Another method may assume that the margin is distributed differently.
The resulting fair probabilities may therefore differ.
For serious analysis, the de-vigging methodology should always be recorded and kept consistent when comparing historical results.
A single bookmaker does not necessarily provide the strongest estimate of market probability.
Suppose several sources offer prices for the same match:
| Source | Home Odds | Draw Odds | Away Odds |
|---|---|---|---|
| Bookmaker A | 2.05 | 3.35 | 3.70 |
| Bookmaker B | 2.10 | 3.30 | 3.60 |
| Bookmaker C | 2.02 | 3.40 | 3.75 |
| Exchange Market | 2.08 | 3.32 | 3.65 |
Each market can be converted into implied probabilities and de-vigged independently.
The resulting estimates can then be compared to determine whether the different sources are broadly agreeing.
A market consensus attempts to summarize information from multiple prices rather than relying on one bookmaker.
A simplified process is:
Multiple Odds Sources → Implied Probabilities → De-Vigging → Consensus Probability
For example, suppose several sources produce the following de-vigged Home probabilities:
| Source | De-Vigged Home Probability |
|---|---|
| Bookmaker A | 47.2% |
| Bookmaker B | 48.1% |
| Bookmaker C | 47.8% |
| Exchange Market | 48.4% |
The estimates are clustered around a similar range.
This provides a stronger indication of market consensus than relying on a single isolated quotation.
De-vigged market probabilities can serve as a benchmark for an independent prediction model.
Suppose:
Market Probability = 48%
Model Probability = 53%
The difference is:
53% − 48% = 5 percentage points
This is a model-market disagreement.
Such disagreement can be useful because it identifies an area where the model and market do not agree. However, it should be treated as a research signal rather than automatic proof of an exploitable edge.
The model could be wrong. The market could contain information that the model has not incorporated. The available odds could also change before any decision is made.
Historical football odds data makes it possible to study de-vigging across large samples rather than relying on individual examples.
Historical datasets containing bookmaker odds can be processed match by match to calculate the implied probability, overround and de-vigged probability for each market.
A researcher can then investigate questions such as:
Large historical samples make it possible to test these questions statistically rather than relying on isolated observations.
De-vigging can also be applied at different stages of a market's life.
An opening market may produce one probability distribution while the closing market produces another.
| Market Stage | De-Vigged Probability |
|---|---|
| Opening | 46.8% |
| Closing | 49.1% |
The movement from 46.8% to 49.1% indicates that the market's normalized assessment changed as new information, trading activity and price discovery occurred.
This is one reason closing prices are useful when evaluating historical market efficiency.
De-vigging can also help put Closing-Line Value (CLV) into a probability framework.
Suppose a selection is available at 2.20 early in the market and later closes at 2.00.
The bettor obtained a better price than the eventual closing market.
By recording the de-vigged probabilities associated with both prices, an analyst can examine how the market's estimated probability changed between entry and closing.
Over a sufficiently large sample, this can help researchers study whether particular pricing signals are associated with favourable subsequent market movement.
Consider the following hypothetical football market:
| Outcome | Odds | Raw Probability | De-Vigged Probability | Fair Odds |
|---|---|---|---|---|
| Home | 2.00 | 50.00% | 47.89% | 2.09 |
| Draw | 3.40 | 29.41% | 28.17% | 3.55 |
| Away | 4.00 | 25.00% | 23.94% | 4.18 |
First, calculate the total implied probability:
50.00% + 29.41% + 25.00% = 104.41%
Next, calculate the overround:
104.41% − 100% = 4.41%
Then normalize each probability:
Home = 50.00 ÷ 104.41 = 47.89%
Draw = 29.41 ÷ 104.41 = 28.17%
Away = 25.00 ÷ 104.41 = 23.94%
Finally, convert the normalized probabilities into fair odds:
Home = 1 ÷ 0.4789 ≈ 2.09
Draw = 1 ÷ 0.2817 ≈ 3.55
Away = 1 ÷ 0.2394 ≈ 4.18
The market has therefore moved from quoted bookmaker prices of 2.00, 3.40 and 4.00 to approximate proportional de-vigged fair prices of 2.09, 3.55 and 4.18.
Now suppose another available price for the Home outcome is:
2.20
The estimated fair price from the de-vigged market is approximately:
2.09
The available price is therefore higher than the estimated fair price.
Using the de-vigged probability:
EV = (0.4789 × 2.20) − 1 ≈ +5.36%
This indicates positive theoretical EV relative to the de-vigged market estimate.
But the conclusion should remain conditional: the result depends on the assumption that the de-vigged market probability is a reasonable estimate of the underlying probability.
De-vigging is particularly useful when:
Market Odds↓
Implied Probabilities
↓
Total Implied Probability
↓
Overround
↓
De-Vigging Method
↓
Normalized Probability
↓
Fair Odds
↓
Market Consensus
↓
Independent Model
↓
Model-Market Difference
↓
Expected Value
↓
Closing-Line Analysis
De-vigging is not about discovering a perfectly accurate "true probability." It is about removing the bookmaker's embedded margin to create a cleaner representation of what the market prices imply.
The basic process is:
Odds → Implied Probability → Overround → De-Vig → Fair Probability → Fair Odds
Once fair odds have been estimated, they can be compared with independent models and currently available prices.
The most important lesson is that fair odds are an analytical benchmark, not a guarantee. The quality of the benchmark depends on the market, the available data and the de-vigging methodology used.
For quantitative sports-betting analysis, this distinction is crucial: the objective is not simply to remove the margin, but to understand what the resulting market probability tells you, how reliable that estimate is and how it compares with independent evidence.
