Ask two books for a price on the same market and you get two numbers that look directly comparable. They are not. Each number carries its book's margin, and margins differ, so the gap between two listed prices mixes a real difference of opinion with a difference in surcharge. Strip each book's margin first, then compare the fair prices that remain.
A worked example
Take a two-way market. Book A lists 2.10 and 1.72. Book B lists 2.05 and 1.78. On listed prices alone, A looks better on the first outcome (2.10 beats 2.05) and B looks better on the second (1.78 beats 1.72).
Convert each price to an implied probability, one divided by the price, and add the pair. Book A's probabilities are 47.62% and 58.14%, a sum of 105.76%, so its overround is 5.76%. Book B's are 48.78% and 56.18%, a sum of 104.96%, an overround of 4.96%. The two books are charging different amounts for the same market.
Remove each margin proportionally, dividing each probability by its book's sum. Book A's fair view is 45.03% and 54.97%, fair prices 2.22 and 1.82. Book B's fair view is 46.48% and 53.53%, fair prices 2.15 and 1.87. The books genuinely disagree: B rates the first outcome about 1.4 percentage points higher than A does. That disagreement was invisible in the listed numbers, because the listed gap of five cents mixed the opinion gap with the margin gap.
Why the listed comparison misleads
A listed price answers two questions at once: what does the book think, and what does the book charge. When the charges differ, ranking books on listed prices rewards the lower margin as often as it rewards the better number. A book that shades its margin on one side can show the best listed price on that side while holding the weaker view of it.
The fair price isolates the opinion. Once both books are reduced to fair probabilities, comparisons become honest: whose number is higher, by how much, and whether either listed price beats the blended fair view. In the example above, the blended fair price on the first outcome is about 2.19, the average of 2.22 and 2.15 in probability terms. Both listed prices, 2.10 and 2.05, sit below it.
Doing the removal in practice
The arithmetic is one line per outcome: implied probability is one divided by the decimal price, the overround is the sum minus one, and the fair probability is each share of the sum. The no-vig calculator does exactly this for a complete set of decimal prices, showing the overround, the normalized probabilities and the fair prices, checked 2026-10-23.
Two cautions from the calculator's own stated assumptions. Proportional removal spreads the margin evenly across outcomes, which is a model, not a fact: uneven markets may carry more margin on the longshot. And the tool converts prices into fair prices; it does not decide whether any resulting number is worth acting on.
Compare like with like
Fair prices only settle a comparison if both sides price the same thing. Before trusting the math, confirm the keys match: the same event, period, market, line and selection, at close timestamps, as the normalization guide lays out. A fair price on the full-match market and a fair price on the first-half market are not two views; they are two markets.
Suspended or stale prices break the comparison quietly. A book that froze its board ten minutes ago is showing you an old opinion, and no amount of margin math fixes a stale input.
A workflow that holds up
For each book you track, on each market you compare:
- Collect the complete set of prices for the market, every mutually exclusive outcome.
- Compute the overround and note it; a book drifting from 103% to 107% is telling you something.
- Remove the margin proportionally and record the fair probabilities.
- Blend the fair probabilities across books into one consensus view.
- Compare listed prices against that consensus, not against each other.
The habit costs a few divisions per market and pays every time you read a feed. Listed prices are for placing; fair prices are for thinking.