No-vig fair prices depend on the removal method you choose. Proportional normalization divides every implied probability by their sum. The power method raises each implied probability to an exponent chosen so the sum lands on exactly one. On balanced prices the two agree. On uneven prices they diverge, because each method puts the margin in a different place.
Two methods, one market
Take a two-way market priced at 1.25 and 4.00. Each price converts to an implied probability of one divided by the price: 0.800 and 0.250. The sum is 1.050, so the overround is 5.0%.
Proportional normalization divides each probability by 1.050. The favorite keeps 76.19% and a fair price of 1.3125. The underdog keeps 23.81% and a fair price of 4.20.
The power method searches for an exponent k that makes the adjusted probabilities sum to one. Here k is about 1.10, because 0.800^1.10 plus 0.250^1.10 equals 1.000. The favorite's fair chance rises to 78.24%, a fair price of 1.28. The underdog falls to 21.76%, a fair price of 4.59.
| Method | Favorite fair price | Underdog fair price |
|---|---|---|
| Proportional | 1.3125 | 4.20 |
| Power, k about 1.10 | 1.28 | 4.59 |
Same inputs, two different fair prices, and the difference is not rounding.
Why uneven prices split the methods
Proportional removal takes the same relative share from every outcome. Each implied probability loses the same percentage of itself, whatever its size.
The power method takes more from the longshot. Raising a small probability to a power above one shrinks it faster than a large one, so the 0.250 side gives up more than the 0.800 side. That shape is not arbitrary. Prices on longshots tend to carry more margin than prices on favorites, a pattern studied often enough to have a name: the favorite-longshot bias. The power method's geometry matches that pattern, which is why some models prefer it when one side is heavily odds-on.
On a balanced market the question never comes up. Two prices of 1.91 each give a sum of 104.71%, and both methods return a fair price of 2.00.
Neither method measures reality
Both are models. They turn posted prices into a consistent no-vig estimate; they do not establish the actual probability of anything. The proportional no-vig calculator on this site states that assumption next to its results, and the same warning applies to the power method and to any other removal rule.
The practical consequence is about consistency. If you judge a bookmaker's price against a proportional fair price on Monday and a power fair price on Tuesday, part of the movement you see is the method, not the market. Pick one method, write it into your code once, and apply it everywhere the same feed is consumed.
Choosing a method for your pipeline
For most pipelines proportional normalization is the right default. It is one division per outcome, needs no solver, reproduces exactly in a spreadsheet or a database view, and matches what the tools on this site return. The power method earns its keep when your work concentrates on uneven markets and you have evidence that the margin leans on the longshot side.
Whatever you choose, convert every incoming price to decimal first so the inputs are comparable: the odds converter handles decimal, American and fractional inputs and shows the implied probability with the margin still in. Run the removal on a complete market only, meaning every mutually exclusive outcome from the same market at the same moment. The calculator accepts up to twelve outcomes and flags an underround rather than normalizing it silently, as stated on its page, checked 2026-09-29.
A fair price is a modeling choice applied to real prices. Settle the method question once, early, and do not re-litigate it on every match.