Work Out What Any Market Costs You Per 100 Staked
A coin flip priced 1.91 on both sides is not fair, and the unfairness has a price.
Overround, margin and cost per 100 staked are three different numbers
On a two-way market priced 1.91 on both sides, overround is the sum of the implied probabilities, 1/1.91 + 1/1.91 = 1.0471, or 104.7%. Margin is the excess over 100%, here 4.71 points, and it is the figure most sites call the margin. Cost per 100 staked is 100 - 100/1.0471 = 100 - 95.50 = 4.50, the average loss per 100 staked if the prices were fair before the bookmaker's cut. So a 4.71% margin does not cost 4.71. This bookmaker margin calculator returns all three for markets of two to six outcomes, in six odds formats.
Put a market through the calculator
Copy the prices of any market, in whatever format the screen shows, and the calculator returns the overround, the margin and the cost per 100 staked, then breaks the market down outcome by outcome. It measures a price; the guide to which bookmaker has the best odds is where the search for a cheaper one starts.
Read the readouts first, then the table: one says what the market costs, the other says where that cost lands.
How to read what a bookmaker margin calculator returns
To calculate a bookmaker margin by hand, add up 1 divided by each decimal price to get T, subtract 1 for the margin, and use 100 - 100/T for the cost per 100 staked. The readouts use the names from the answer block, and cost per 100 staked is the figure to compare between markets. The table adds, for each outcome, the implied probability, the fair odds and the cost per 100 staked under three assumptions about where the margin sits: Proportional, Additive and Power. Proportional costs the same on every outcome by construction, which is why the third readout is a single number.
Tick the box to compare two books, or a bookmaker and an exchange. Each outcome takes a second price, and the readouts become the cost at A, the cost at B and the sum of the inverses of the best prices combined. A sum under 100% is arithmetic on a snapshot that promises no gain, since prices move within seconds and accounts can be limited; the guide to prices that combine under 100% covers that case. Betting is for adults aged 18 or over, and legality varies by country.
For an exchange, type its commission rate on net winnings into the field for that book. At 2.02 on both sides with a 5% rate, each price becomes 1 + 1.02 × 0.95 = 1.969, the inverses sum to 1.0157, the margin is 1.57% and the cost is 1.55 per 100 staked. The rate is yours to enter, and back and lay betting shows how the commission an exchange charges cuts a back price.
Prices adding up to 100% or less leave the Power method nothing to remove, and a heavy margin can push a long price's Additive fair probability to zero or below. In both cases the table reads n/a, because a method's assumption has stopped working.
Same margin, three different splits
This is why the table has three columns. Here is a three-way market priced 1.60, 4.00 and 6.50 (fictional prices). The implied probabilities, 62.50%, 25.00% and 15.38%, add up to 102.88%, so the margin is 2.88% and the cost is 100 - 100/1.0288 = 100 - 97.20 = 2.80 per 100 staked. That figure is exact if you back every outcome so the same amount returns whichever one wins: stakes of about 62.50, 25.00 and 15.38 total 102.88 and return 100. Bet on a single outcome and the cost depends on how the bookmaker spread its margin, which you cannot see.
Three common assumptions fill the gap. The proportional method takes the same share off every probability, so each outcome costs 2.80. The power method raises every probability to one exponent, about 1.03 here, until they total 100%, and gives 1.48, 4.30 and 5.76 per 100 staked. The additive method removes the same 0.96 points from every probability and gives 1.54, 3.85 and 6.25. Both put less cost on the favourite and more on the long price. On 1.91 and 1.91 all three agree at 4.50.
Cost per 100 staked on 1.60, 4.00 and 6.50, by method
Illustrative dataView the data as a table
| Outcome | Proportional | Power | Additive |
|---|---|---|---|
| Price 1.60 | 2.80 | 1.48 | 1.54 |
| Price 4.00 | 2.80 | 4.30 | 3.85 |
| Price 6.50 | 2.80 | 5.76 | 6.25 |
The bars show how far apart the methods land, and they cannot say which one is right.
Is margin the same as overround?
Not quite, and this is where the 4.71 goes wrong. On 1.91 both sides the overround is 104.71% and the margin, as most sites use the word, is the 4.71 points above 100. The cut of the money staked is smaller, because the base is larger: to get 100 back whichever side wins you stake 104.71, and the 4.71 you lose is 4.50% of that. Some writers keep the word margin for that 4.50, so check which one a page means.
On a 2% market the margin and the cost differ by 0.04 points; at 30% they differ by nearly 7.
| Margin | Overround | Cost per 100 staked | Expected cost over 1,000 bets of 100 |
|---|---|---|---|
| 2% | 102% | 1.96 | 1,961 |
| 4.71% | 104.71% | 4.50 | 4,498 |
| 8% | 108% | 7.41 | 7,407 |
| 30% | 130% | 23.08 | 23,077 |
Cost per 100 staked = 100 - 100/T, where T is the overround as a fraction, that is 1 plus the margin. The last column uses the unrounded cost, so it can differ by a unit from 1,000 times the rounded figure.
These are averages: one run of 1,000 bets can end well above or below the 4,498, and the 30% row is a computed illustration for a wide market such as an outright with 20 runners, not a quote. A lower margin lowers the expected cost and nothing more; it makes no bet win. Stake 1,000 a week with nobody limiting you, and a cost of 1.96 per 100 means about 19.6 a week in expectation against 74.1 at 7.41, a gap decided only by the market you chose.
A bonus rollover is margin too: staking 1,000 on markets like 1.91 both sides costs 45 in expectation (1,000 × 4.50%) before a bet has won or lost, so the rollover attached to a bonus deserves the same arithmetic.
One price, six ways to write it
The same price changes shape from one screen to the next. Decimal, fractional and American odds are the familiar three; the other three need a rule. Hong Kong odds are the net profit per 1 staked, which is the decimal price minus 1. Malay and Indonesian odds are signed variants of it that some bookmaker and broker screens use.
From a decimal price d, Hong Kong is d - 1. Malay is Hong Kong when that is 1 or less, otherwise -1 divided by Hong Kong. Indonesian is Hong Kong when that is 1 or more, otherwise -1 divided by Hong Kong. In both, a positive number is the profit on a stake of 1 and a negative number is the stake you must risk to win 1.
| Decimal | American | Fractional | Hong Kong | Malay | Indonesian |
|---|---|---|---|---|---|
| 1.909 | -110 | 10/11 | 0.909 | 0.909 | -1.10 |
| 3.00 | +200 | 2/1 | 2.00 | -0.50 | +2.00 |
| 2.00 | +100 | 1/1 | 1.00 | +1.00 or -1.00 | +1.00 |
| 1.50 | -200 | 1/2 | 0.50 | 0.50 | -2.00 |
Watch the minus sign: it marks a price below 2.00 in American and Indonesian odds and above 2.00 in Malay odds, so -110 and -1.10 are the same 1.909 while -0.50 is a 3.00 outsider. At exactly 2.00, sources disagree on whether Malay reads +1.00 or -1.00; the calculator accepts either and shows +1.00.
Each calculator row has its own format selector. Switching it rewrites the price for the same odds, and a line under the price shows the decimal equivalent (the quickest way to convert a single price). Fractions are typed as 10/11 or evens, and American odds must be +100 or above, or -100 or below.
Does the formula match what bettors actually lose?
The cost formula is arithmetic; whether real prices behave that way is another question. A 2025 paper by Tadgh Hegarty and Karl Whelan of University College Dublin tests it on 151,683 European football matches from 2005/06 to 2024/25, with odds averaged across bookmakers, using data made available by Joseph Buchdahl. It compares each season's predicted average loss with the loss that a 1 unit bet on every home win, draw and away win actually suffered.
Football: average loss per bet, predicted and realised, by season
Sourced dataView the data as a table
| Season | Predicted by the overround formula | Realised |
|---|---|---|
| 2005/06 | 10.0% | 11.9% |
| 2006/07 | 9.7% | 11.4% |
| 2007/08 | 9.2% | 11.1% |
| 2008/09 | 8.5% | 10.2% |
| 2009/10 | 8.0% | 10.1% |
| 2010/11 | 7.6% | 8.9% |
| 2011/12 | 7.5% | 9.3% |
| 2012/13 | 7.0% | 7.7% |
| 2013/14 | 6.9% | 8.6% |
| 2014/15 | 6.6% | 8.1% |
| 2015/16 | 6.6% | 7.7% |
| 2016/17 | 6.6% | 8.1% |
| 2017/18 | 6.4% | 8.5% |
| 2018/19 | 6.4% | 8.5% |
| 2019/20 | 6.0% | 7.4% |
| 2020/21 | 6.0% | 6.3% |
| 2021/22 | 5.9% | 7.0% |
| 2022/23 | 5.7% | 7.7% |
| 2023/24 | 5.8% | 8.1% |
| 2024/25 | 5.8% | 7.7% |
The realised loss sits above the predicted one in all 20 seasons, by as little as 0.3 points in 2020/21 and as much as 2.3 in 2023/24. Over the whole sample the authors report 7.1% predicted against 8.7% realised; in tennis, on 130,243 ATP and WTA matches from 2010 to 2024, 5.4% against 7.5%, higher in every one of the 15 years.
Sorted into ten groups by odds, football bets with the shortest odds lost 3% on average and those with the longest 19%; in tennis, 2.5% and 25%. That uneven cost is called the favourite-longshot bias. The proportional column cannot show it, and the power and additive columns lean the same way on the example above, though neither is a measurement.
For Pinnacle, whose access rules are covered in how to bet on Pinnacle Sportsbook, the authors find losses significantly lower than at the other bookmakers in their sample, yet realised losses there also exceed the predicted ones in the 1X2 market. In a footnote, the authors report that the pattern does not apply to Asian handicap betting on football, a market they call popular with professional bettors. It is one footnote about one market, so do not stretch it to every book or every line.
Four reservations before you quote any of this:
- The loss is an equally weighted average over every outcome offered (home, draw and away), not the loss on stakes placed; the authors note that volumes are not public.
- The odds are averages across several bookmakers, not the price of one book.
- The samples are historical: football from 2005/06 to 2024/25, tennis from 2010 to 2024.
- This is an academic paper, a revised draft from April 2025, not operator data.
Read the three cost columns as a range of assumptions, and remember that on this evidence long prices tend to cost more than short ones.
Six terms you will meet on any odds page
- Overround
- The sum of the implied probabilities of every outcome, above 100% when a margin is built in: 104.7% on 1.91 both sides.
- Margin
- The overround minus 100, in points: 4.71 on 1.91 both sides. Some writers use the word for the cost per 100 staked instead.
- Vig or juice
- Slang for the bookmaker's built-in charge, the margin under another name: a two-way line at -110 on both sides, 1.909 in decimal, carries a margin of 4.76, a shade above the 4.71 of 1.91 because 1.91 is rounded up.
- Fair odds
- The price an outcome would carry with the margin removed, 1 divided by its fair probability. It is an estimate, since it depends on the method used to remove the margin.
- Implied probability
- The chance a price suggests, 1 divided by the decimal odds: 1.91 implies 52.36%.
- Hold
- The share of the money staked that a bookmaker keeps. With stakes balanced so the same amount is paid on every outcome, it equals the cost per 100 staked written as a percentage: 4.50% on 1.91 both sides.
Once you know the cost, where next?
Knowing what a price costs does not tell you whether it is worth taking, and value betting starts from the fair odds this page produces. If you want to compare venues rather than single markets, what makes a bookmaker sharp is the next read.