Methods to remove bookmaker margin from odds
PuntPilot offers four methods for removing the vig from bookmaker odds: additive, multiplicative, power, and shin.
We'll briefly describe each method below. By default, PuntPilot uses the power method, as it tends to give the most accurate results when backtested across a wide range of markets. We do not recommend changing this unless you know what you're doing.
The additive method removes the bookmaker’s margin by subtracting the exact same amount from every implied probability.
For example, if three runners are listed at 45%, 35%, and 30% (total = 110%), the extra 10% is split evenly, so each loses ≈ 3.33%.
After adjustment, the probabilities become roughly 41.7%, 31.7%, and 26.7%. It’s easy to apply and easy to explain, but because every outcome gets the same deduction, long-shots can be pushed unrealistically low or even negative in larger markets.
With the multiplicative approach, you remove the margin by scaling all probabilities down proportionally until they sum to 100%.
Using the same example (45%, 35%, 30% = 110%), each value is divided by 1.10, giving new probabilities of roughly 40.9%, 31.8%, and 27.3%.
This keeps the same relative price shape and avoids negative values. However, it assumes the margin is evenly spread across all outcomes, which isn’t always true in real betting markets where favourites and long-shots are often priced differently.
The Shin method assumes that some portion of betting comes from well-informed bettors who favour the most likely outcome.
This creates a price imbalance in market behaviour, so the method mathematically adjusts the probabilities so long-shots get a comparatively bigger correction than favourites.
Using the earlier numbers, this method might move the 45% favourite only slightly, while pushing the 30% outsider down more significantly once the model estimates how much “insider activity” exists. It typically produces more realistic fair odds, especially in racing markets.
The power method adjusts implied probabilities by raising each to a chosen exponent, which reduces larger probabilities and boosts smaller ones before normalising the total to 100%.
In the same 45%/35%/30% example, a calculated exponent might reduce the favourite more gently while increasing the outsider slightly, leading to a smoother redistribution than the Shin or additive approaches.
This method is mathematically stable, avoids negative values, and has been shown to produce strong real-world accuracy across several sports and racing datasets.