Removing the bookmaker margin is a useful first step when you want to compare market probabilities. The calculation turns a set of prices whose implied probabilities exceed 100% into a set that sums to 100%. It does not reveal the true probability of each outcome. That distinction matters whenever a historical analysis uses “fair odds” as its benchmark.
Start with a complete market
Consider a hypothetical football 1X2 market: home win 2.20, draw 3.40 and away win 3.30. These selections are mutually exclusive and cover every result for the defined match period. Keep all three prices from the same bookmaker and observation time. Using the best price from a different bookmaker for each outcome would describe a different, synthetic market.
Convert each decimal price to a raw implied probability by taking its reciprocal. The home price implies 45.4545%, the draw 29.4118% and the away price 30.3030%. The total is 105.1693%. Subtracting 100% gives an overround of 5.1693 percentage points.
Pinnacle's margin explainer describes this whole-market calculation. The term margin is often used loosely: overround is a property of the quoted prices, not the bookmaker's realised profit on the match.
Apply proportional normalisation
The simplest transparent approach divides each raw probability by their total. In decimal notation, let S = 1/2.20 + 1/3.40 + 1/3.30 = 1.051693. The fair-probability estimate for an outcome is then (1/odds) ÷ S. This assumes the overround is allocated proportionally across the raw implied probabilities.
- Home: 0.454545 ÷ 1.051693 = 43.2203%.
- Draw: 0.294118 ÷ 1.051693 = 27.9661%.
- Away: 0.303030 ÷ 1.051693 = 28.8136%.
These estimates add to 100%, subject to rounding. Convert them back to decimal fair odds by taking their reciprocals: approximately 2.3137, 3.5758 and 3.4706. Retain full precision for calculations and round only the displayed result.
Why subtracting the overround equally gives another answer
An additive approach would subtract one third of the 5.1693 percentage-point excess from each outcome. That produces approximately 43.7314%, 27.6887% and 28.5799%. The method also produces a 100% book, but the home estimate differs from proportional normalisation by about half a percentage point.
Neither method becomes correct merely because its probabilities sum to 100%. Each makes an assumption about how margin is distributed. Equal subtraction can even produce negative probabilities for sufficiently unlikely outcomes in some markets. More complex approaches exist, but complexity does not replace checking assumptions and performance on fresh data.
Use fair odds as a declared reference
Suppose another bookmaker offers 2.40 for the home win in this example. Against the proportional benchmark, the estimated expected return is 0.432203 × 2.40 − 1 = 3.73%. That figure is conditional on the benchmark being a good probability estimate. A different margin-removal method, a stale reference quote or new team news could change it.
This is why a comparison report should name the bookmaker, timestamp, full price set and method. Writing “3.73% value” without those assumptions makes an estimate appear more certain than the evidence supports.
Keep overround separate from payout and actual results
For this price set, 1 ÷ S is approximately 95.0847%. The complementary figure, 4.9153%, is sometimes used as a normalised margin under a proportional model. It is numerically different from the 5.1693% overround. Always state which measure a chart uses.
Actual betting returns also depend on the selected outcomes, available stakes, settlement and underlying probabilities. An observed losing run cannot establish the bookmaker's margin, and a small overround does not promise a profitable strategy.
Use this calculation alongside consistent bookmaker comparisons and our closing line value example. For historical work, keep the same margin-removal method across the sample and test whether conclusions survive a reasonable alternative.