Fair Odds and No-Vig Math: The Calculation Behind Every Decision

This is the most basic operation in price analysis, and the one most people skip. Without it, every comparison between a model estimate and a market price is wrong in a consistent direction — always against the model.

Step 1: probability to fair odds

fair odds = 1 / probability

Probability	Fair odds
75%	1.33
60%	1.67
54%	1.85
50%	2.00
45%	2.22
33%	3.03
25%	4.00

Fair odds are one divided by the probability. A 54% view is fair at 1.85. A price below that implies the book is more confident than the model. A price above it is where the conversation starts. Fair odds alone are not a recommendation — they are only the probability expressed as a price.

Step 2: the displayed price is not a probability

This is where most analysis goes wrong.

Take a match priced at 2.10 / 3.40 / 3.60.

1 / 2.10 = 47.62%
1 / 3.40 = 29.41%
1 / 3.60 = 27.78%
total    = 104.81%

Real probabilities sum to 100.00%. The 4.81 points of excess are the overround — what the book keeps on this market, baked into all three prices.

The practical consequence: compare a 46% model view to 47.62% and you conclude the market is more confident than you. After normalisation the market implies 45.44%, and it is you who is the more confident party. Same inputs, opposite conclusion.

Step 3a: proportional devigging

Divide each implied probability by the total.

47.62 / 104.81 = 45.44%
29.41 / 104.81 = 28.06%
27.78 / 104.81 = 26.50%
                 -------
                 100.00%

Advantages: one line, no parameters, always works. Disadvantage: it assumes the margin is spread proportionally across outcomes, which is probably not true.

Step 3b: the power method

Instead of dividing, find an exponent k such that:

Σ (implied probability)^k = 1

Solve for k by bisection, then normalise. Because exponentiation compresses small numbers more sharply than large ones, this removes more margin from favourites and less from longshots.

Why that is interesting: there is a well-documented favourite-longshot bias in betting markets, where longshots are priced shorter than their true probability and favourites longer. The power method partially corrects for it.

When the difference matters

Situation	Difference between methods
Balanced match, tight margin	Negligible — under a probability point
Heavy favourite vs longshot	Noticeable — can exceed a point on the longshot
High-margin market	Large — more margin to distribute means the distribution matters
Two-way market	Small — fewer outcomes, less room for bias

Worked example. A match priced 1.30 / 5.50 / 11.00 implies 76.92%, 18.18% and 9.09%, totalling 104.19%. Proportional devigging gives 73.83% / 17.45% / 8.72%. The power method leaves the longshot slightly higher, around 8.9%, and pulls the favourite slightly lower. Two tenths of a point sounds trivial — but relative to a 9% probability it is roughly 2% of the value, and anyone hunting edges on longshots needs to know which method they are measuring with.

Step 4: measure the edge in points, not percent

edge = model probability − no-vig market probability

If the model says 51% and the clean market says 45.44%, the edge is 5.6 percentage points. Not 12%. Expressing it as a relative percentage inflates the perceived size by more than double and leads people to act on gaps that sit inside their own error bars.

Step 5: the minimum acceptable price

Fair odds assume the probability is exact. It never is. So before inverting, add an uncertainty budget:

minimum acceptable price = 1 / (model probability − uncertainty budget)

The budget is not fixed. It grows with:

• Unconfirmed lineups, especially close to kickoff.
• Low data quality — short history, few price sources, no xG.
• Model disagreement between ensemble components.
• Market volatility across recent price snapshots.

Worked example: model probability 51%, uncertainty budget 3 points.

fair odds                = 1 / 0.51 = 1.96
minimum acceptable price = 1 / 0.48 = 2.08

If the market offers 2.02, there is an edge against fair odds and no edge against the uncertainty. That is a PASS.

Step 6: the price zone and a real expiry

From all of the above comes a simple reading of the current price:

Zone	Meaning
Green	The price still supports the thesis
Watch	Thin margin — sensitive to small movement
Wait	An information event is coming, usually lineups
Red	The price no longer supports the thesis
Expired	The estimate is no longer valid

Each carries a genuine expiry: the next scheduled price refresh, or five minutes before kickoff, whichever comes first. No manufactured countdowns. If a timer cannot be explained, it is not shown.

Three mistakes that repeat

Devigging a partial market. You cannot normalise one outcome. You need every outcome in that market — three for 1X2, two for over/under. Taking a single price and dividing it by some constant is not devigging, it is invention.

Mixing sources within one market. Taking the home price from one book and the draw from another and then normalising produces an imaginary negative margin. Normalisation must run on all outcomes from the same source at the same moment.

Treating the cleaned number as exact. After normalisation you get 45.44%, which looks precise. It is not. It depends on the method, the source and the moment of the snapshot. A one-point gap against that number is not an edge, it is noise.

Why low margin is not the same as a good price

A tight margin means the book is confident, usually because there is a lot of information and a lot of money on the market. That cuts both ways: the price is closer to the true probability, and it is harder to find a gap in it.

The reverse also holds. A high margin on a marginal league signals that the book is protecting itself against its own ignorance — and in those same leagues our data quality tends to be weakest. Two uncertain parties do not create an opportunity; they create doubled noise. An edge measured against a 9% margin needs to be substantially larger than one measured against 3%, because a meaningful share of it is devigging error.

The full sequence

1. Take all outcome prices from one reference source at one moment.
2. Convert to implied probabilities.
3. Normalise — proportional or power.
4. Compare to the model probability, in percentage points.
5. Require the gap to clear the fixture's uncertainty budget.
6. Compare fair odds and the minimum acceptable price to the price you can actually get.

Steps 1–4 are a calculator. Steps 5–6 are what turn arithmetic into a decision, and they are set out in full on the fair price page and the decision states page.

---

18+. WinPIQ is an analysis tool, not advice and not a promise. Betting can be addictive and money can be lost. Only stake what you can afford to lose, and if betting stops being entertainment, seek help. WinPIQ is not affiliated with Winner or the Israeli Council for the Regulation of Sports Betting.