Two words get used as if they were interchangeable: overround and hold. They are not. One is a property of the prices in front of you, computable in a spreadsheet cell. The other is a property of what happened after the bets settled, and no odds feed will ever show it to you. Mixing them up produces two specific mistakes, and both are avoidable.
Overround is a property of the prices
Convert every price in a complete market to an implied probability, add them up, and subtract one. That excess is the overround. Two sides at 1.91 each sum to 104.71%, so the overround is 4.71%, the same worked figure the no-vig calculator shows for that pair, checked 2026-10-26.
Everything about the number comes from the prices alone. No stakes, no results, no book internals. That is what makes it computable from a feed, and it is also the limit of what it can tell you.
Hold is a property of the outcomes
Hold is what the book actually kept: the share of all stakes that remained with the operator after the winning bets were paid. That number needs two things the prices never carry: how the stakes were distributed across the outcomes, and which outcome won.
A balanced book priced at a 4.71% overround expects to keep close to that percentage when stakes line up with the probabilities. A lopsided book, where the crowd piled onto one side, can hold far more on a quiet night and far less, even negative, when the popular side wins at a bad price. Same overround, wildly different holds. The overround is the toll booth's rate; the hold is the day's takings, and the day's traffic decides that.
Mistake one: reading profitability into a price
The first mix-up treats a market's overround as the operator's expected profit on that market. It is not. Overround sets the terms of each bet; hold is decided by volume and results. A feed reader can legitimately say a market is expensive or cheap to trade in. Nothing in the feed says whether the operator made money on it, and any sentence that derives the second from the first alone is overreach.
Mistake two: feeding hold into margin removal
The second mix-up runs the other way. Margin removal, the math that turns listed prices into fair ones, consumes the overround: the excess over 100% is the material being redistributed. Someone who substitutes a published hold figure, a hold they estimated, or a half-remembered average into that role is removing the wrong quantity. The fair prices that fall out will be wrong in proportion to the confusion.
The converter and the no-vig calculator both work from prices for the same reason: prices are what a data feed can actually provide, checked 2026-10-26.
The reading rule
Keep the words sorted by their inputs. If a number was computed from prices alone, it is an overround-style figure, whatever its label. If it claims to say what an operator kept, it came from settlement data you cannot see, and the honest response is to ask for its source.
From an odds feed you will only ever compute the first kind. That is not a limitation to work around; it is the correct scope of the data, and it is exactly the input fair pricing needs.