How betting odds work: implied probability and the bookmaker's margin

Five minutes before Türkiye played Italy in the Nations League on Monday 28 September, two bookmakers were pricing the same match. Pinnacle's prices carried a margin of 3.3%. William Hill's carried 10.1%, three times as much. Both sets of odds look ordinary. This guide shows you how to read a price, find the fee in it and take the fee out, using that one match from our odds database, which recorded its prices at 16 bookmakers and the Betfair exchange.

What do betting odds mean? Decimal, fractional and American#

Odds tell you two things at once: how much a winning bet pays, and how likely the bookmaker thinks the outcome is. The same price can be written three ways, and all three pay out exactly the same.

Decimal odds (2.58) are the total you get back for every €1 you stake (the amount you bet), the stake included. €10 at 2.58 returns €25.80: €15.80 profit plus your €10. Most of Europe, including Estonia, uses decimals, and they are the easiest to compare.

Fractional odds(6/4) are the profit for each unit staked, without the stake. 6/4 means €6 profit for every €4, so €10 returns €25. To convert: decimal = fraction + 1. "Evens" (1/1) is 2.00. Fractions are traditional in the UK and Ireland.

American odds (+150 or −200) are built around 100. A plus number is the profit on a 100 stake: +150 wins 150. A minus number is what you must stake to win 100: at −200 you stake 200. Plus means underdog, minus means favourite.

Implied probability formula (decimal odds)#

Every price hides a probability. Divide 1 by the decimal odds and you get the implied probability: the chance the price says the outcome has, before the bookmaker's fee is taken out.

The same number is your break-even rate. A bet at 2.58 has to win 38.8% of the time just to get your money back over many bets. Win less often at that price and you lose money, however good the win rate sounds. That is how a tipster with a 76% win rate can still lose: in our Betzena review, its Over 1.5 goals picks, at an average price of 1.28, needed to win about 79% of the time (each price's break-even, averaged).

What is the bookmaker's margin?#

A football match has three outcomes: home win, draw, away win. That market is called 1X2. Turn all three prices into implied probabilities and add them up. If the prices were fair, they would total exactly 100%, because one of the three must happen. They never do. The amount over 100% is the overround, which most people call the margin (or the vig). It is how the bookmaker gets paid: each price is a little shorter (lower, paying less) than the real chance would justify.

The average bookmaker margin differs a lot from one bookmaker to the next, and from one league or market to another. The same match can cost you about three times as much in one place as in another, as the next example shows.

Türkiye v Italy: the same match at two bookmakers

Here is the match from our database: UEFA Nations League, Monday 28 September 2026, kick-off 20:45 CEST at the Bursa Büyükşehir Stadyumu in Bursa. These are each book's last prices before kick-off, both read at 20:40. Both books had moved their prices in the previous ten minutes, so neither was out of date. Pinnacle is a sharp bookmaker: it takes big bets from skilled bettors and runs low margins. William Hill is a high-street (or soft) bookmaker: it serves recreational customers and charges a higher margin.

William Hill was shorter on all three outcomes. A lower price on the same outcome is simply a more expensive bet: you pay more for the same chance. Most of its extra margin sat on the two teams rather than the draw.

Overround vs hold: what you actually pay

A 10.1% overround does not mean you lose €10.10 of every €100. If the book balanced its bets perfectly, it would keep about €9 of every €100 bet with it, whoever won (1 − 1 ÷ 1.101). That cut is called the hold. At Pinnacle's prices it is about €3. You pay it on every bet, win or lose, because it is built into the price.

Was that one match typical? We took the 42 Nations League matches from 24 to 29 September where both books had a price in the last 30 minutes before kick-off. Pinnacle's median margin was 3.8%, William Hill's 10.0%. William Hill charged more on every one of those matches. Margins differ by league and by market; this is one competition in one week.

Fair odds: taking the margin out#

Fair odds are the prices with the margin removed: what each outcome would pay if the implied probabilities added up to exactly 100%. They are an estimate of the real chance, and every other price can be measured against them.

The simple way is proportional: divide each implied probability by the total. Italy's 38.8% becomes about 37.5% (38.76 ÷ 103.33, before rounding). It is easy, and slightly wrong. Bookmakers do not spread their margin evenly: they tend to load more of it onto longshots (unlikely outcomes with long, high prices), a pattern known as the favourite–longshot bias. Proportional removal leaves longshots looking a little too likely and favourites a little too unlikely.

So for three-way markets like 1X2 we use Shin's method, which takes more of the margin off the longer prices, and for two-way markets the power method. Both choices follow published comparisons (see Sources).

With fair odds you can price any bet. Fair chance × odds × stake is the bet's expected return: what it pays back on average over many bets like it. Above your stake is positive expected value; below is negative. William Hill's Italy at 2.40 against a fair 37.6%: 2.40 × 37.6% × €100 = €90.27. Its Türkiye price returned €90.30 and its draw €92.14. Pinnacle's own prices returned €96 to €97. None reaches €100: at a bookmaker's price you start behind. These figures assume Pinnacle's margin-free price is right, which is a strong estimate, not a certainty.

Draw no bet and double chance explained#

These two markets are built from the same three outcomes, so once you have fair chances you can work out their fair prices yourself.

Draw no bet takes the draw out: if the match is drawn, you get your stake back. The fair chance of Türkiye is its share of the non-draw outcomes: 34.5 ÷ (34.5 + 37.6) = 47.8%, fair odds 2.09. Italy gets 52.2%, fair odds 1.92. Pinnacle offers the same bet as Asian handicap 0.

Double chance covers two outcomes with one bet. Türkiye or draw (1X) is 34.5 + 27.9 = 62.4%, fair odds 1.60. Draw or Italy (X2) is 65.5%, fair odds 1.53. Either team to win (12) is 72.1%, fair odds 1.39.

These markets carry a margin too, and it is not spread evenly. William Hill quoted 1.85 on both sides of draw no bet, and those two prices had not moved when its 1X2 prices did, so they may simply have been out of date. Against the fair odds, its Italy side returned about €97 per €100 and its Türkiye side about €88. A price that lags the rest of the book can sit near fair by accident: worth checking, not trusting.

You can also build draw no bet yourself from the 1X2 prices, by putting part of your stake on the draw to cover it. Here William Hill's ready-made 1.85 was the better price on both sides. On another match it can be the other way round, so check both.

What this means for you#

Odds are a probability plus a fee. Once you can turn any price into a percentage, find the margin and compare it with a fair price, you can see what a bet costs before you place it, and spot when one book charges about three times as much as another for the same match. What you cannot do is make the fee disappear: because every price carries a margin, most people who bet lose money over time. A price is good or bad before kick-off. Italy winning 4–1 did not make any of these prices better or worse.

This is the yardstick we use on ourselves. Every pick we publish is sealed with its price before kick-off and then measured against the fair version of the closing price (the last price before kick-off), win or lose. How that works is on our methodology page, any sealed pick can be checked on /verify, and the full record, losses included, is on our track record.

FAQ#

Sources#