Odds & Margins
Low odds or high odds: where the bookmaker keeps more
Across 12,459 matches in the five big leagues, backing every outcome priced between 5 and 10 lost 13.3% of the stake; below odds of 1.50, 1.7%. The margin does not weigh the same on every price.
Published 25 October 2026
The bookmaker keeps far more on high odds than on low ones. Across the 12,459 Premier League, LaLiga, Bundesliga, Serie A and Ligue 1 matches from 2019-20 to 2025-26, backing every outcome at the market's average closing odds would have lost 1.7% of the stake below odds of 1.50 and 13.3% between 5 and 10.
The reason is the gap between what the odds say and what happens. With the margin removed, an outcome priced between 5 and 10 carries an average probability of 15.0%, and it happened 13.8% of the time. Favourites go the other way: below 1.50 the margin-free odds say 72.5%, and the real frequency was 74.4%.
The measurement
| Odds | Bets | Average loss | Margin-free probability | Happened |
|---|---|---|---|---|
| below 1.50 | 2,544 | 1.7% | 72.5% | 74.4% |
| 1.50 to 2 | 4,429 | 2.4% | 55.1% | 56.4% |
| 2 to 3 | 7,684 | 4.5% | 39.1% | 39.2% |
| 3 to 5 | 16,642 | 5.9% | 26.1% | 25.7% |
| 5 to 10 | 5,081 | 13.3% | 15.0% | 13.8% |
| above 10 | 997 | 4.5% | 7.2% | 7.1% |
Every outcome of every match counts as a one-unit bet at the average closing odds: three bets per match, 37,377 in total. «Average loss» is the share of the stake kept by the bookmaker; «margin-free probability» is what the odds imply once the margin is removed; «happened» is how often that outcome actually occurred.
Read the row above 10 with caution. It rests on 997 bets at very high odds, where a handful of wins more or less moves the total: resampling those bets, the average result ranges from an 18.2% gain to a 26.0% loss. That 4.5% is an indication, not a measurement. The other rows rest on thousands of bets and are stable.
Why the cost is not the same at every price
The average margin on the final result in these matches was 4.73%. If the bookmaker spread it evenly, every odds band would cost roughly that. It does not: low odds cost less than the average, and odds between 5 and 10 almost three times as much.
Betting research has a name for this pattern, the favourite-longshot bias, and it has been documented for decades across different sports. The most cited explanation is that bettors prefer high odds, and prices adjust to that demand. We have not tested that: what we measured is the table above.
In Serie A
In Serie A alone there are 2,660 matches and the pattern is the same, steeper at the long end:
| Odds | Bets | Average loss | Margin-free probability | Happened |
|---|---|---|---|---|
| below 1.50 | 572 | 2.1% | 71.2% | 73.1% |
| 1.50 to 2 | 956 | −0.6% | 55.3% | 58.3% |
| 2 to 3 | 1,584 | 2.5% | 38.9% | 39.8% |
| 3 to 5 | 3,519 | 8.0% | 26.1% | 25.3% |
| 5 to 10 | 1,170 | 18.2% | 15.0% | 13.1% |
| above 10 | 179 | 15.8% | 7.6% | 6.7% |
The −0.6% between 1.50 and 2 means that in that band the stake, overall, came back with a small surplus. It is not a figure to rely on: over 956 bets the interval runs from a 5.0% loss to a 6.6% gain, so it comfortably includes both zero and an ordinary loss. With a single league the numbers swing more, which is why the main table puts all five together.
What it means in practice
The margin you read on a bookmaker is an average over the three outcomes, but within the same match it does not weigh the same: the largest share sits on the least likely outcome. For the same stated margin, betting high odds costs more than betting low odds.
High odds also carry huge variance: even a record losing 13% can look positive for months, because a single win at 8 covers many lost bets. The real price only shows over thousands of bets, as in this table.
From the feed, right now
How much each operator keeps right now on the final result, across the whole football list:
The feeds are unreachable right now. We'd rather not show you a number that isn't the current one.
How we calculated it
The odds are the average closing prices of the operators tracked by football-data.co.uk, which publishes results and odds for the main European leagues free of charge: for each match, the average of the prices offered on the three outcomes shortly before kick-off. They are not the odds of a single Italian operator: our own archive starts in July 2026 and is still too young to measure results over thousands of matches.
For each outcome we compute the implied probability (1 divided by the odds), scale it back to 100% by dividing by the sum of the three outcomes to remove the margin, and compare it with what happened. The return is the payout minus the stake, divided by the stake, averaged over every bet in the band. The matches are those of the 2019-20, 2020-21, 2021-22, 2022-23, 2023-24, 2024-25 and 2025-26 seasons, the only ones for which the source reports closing odds.
The margin explained in full, worked through on one match: What the bookmaker's margin is, and how to measure it on a 1X2
Why high odds take much longer to show how things are really going: Yield and variance: how many bets you need before you know