Bookmaker margin microscope
A complete market converts quoted odds into implied probabilities whose sum normally exceeds 100%. Removing that excess does not recover a uniquely determined hidden truth: each de-vig method makes a different assumption about how the bookmaker distributed the margin.
The same quoted odds remain fixed while four legitimate models disagree. Every result below is an estimate under an explicit assumption. None is labeled the true probability.
How the quoted book is composed
One market, four margin-removal assumptions
Each line spans the smallest to largest estimate across the valid methods. The dots answer the same question under different assumptions; their separation is model uncertainty, not rounding error.
What each method assumes
Every implied probability is divided by the same total. Margin is removed in proportion to the quotation.
The same number of percentage points is removed from every outcome. It can become invalid when a longshot's implied probability is smaller than its equal share of the margin.
Every implied probability is raised to one common exponent chosen so that the estimates sum to 100%. It changes favorites and longshots by different relative amounts.
A model derived from a market containing informed and uninformed money. Its fitted parameter redistributes the margin rather than treating every outcome identically.
The balanced book
| Outcome | Quoted odd | Accepted stake | Payout if it wins | Bookmaker profit |
|---|
Margin and vigorish are related but not identical. Margin is the amount above 100% in the implied probabilities. Vigorish here is profit divided by the total money accepted: margin / (1 + margin).
Interpretation: de-vigging removes an internally chosen model of the margin; it does not discover the actual probability of an event. Comparing several bookmakers and evaluating independent information may improve an estimate, but agreement between de-vig methods is not proof of an edge.