How Football Betting Odds Work
Learn how decimal, fractional and American odds express the same price in different formats, how those prices map to raw implied probability, and why a bookmaker market can sum to more than 100%.
Quick conversion table
| Decimal | Fractional | American | Implied prob |
|---|---|---|---|
| 1.20 | 1/5 | −500 | 83.3% |
| 1.50 | 1/2 | −200 | 66.7% |
| 2.00 | 1/1 (evens) | +100 | 50.0% |
| 2.50 | 3/2 | +150 | 40.0% |
| 3.00 | 2/1 | +200 | 33.3% |
| 4.00 | 3/1 | +300 | 25.0% |
| 6.00 | 5/1 | +500 | 16.7% |
| 10.00 | 9/1 | +900 | 10.0% |
One formula: raw implied probability = 1 / decimal odds. Multiply by 100 to express the result as a percentage. This is the break-even win rate at the quoted price, before any market-level margin removal.
Need a conversion tool? The Implied Probability Calculator converts one quoted price between all three odds formats. It does not remove bookmaker margin.
Three formats explained (decimal, fractional, American)
Decimal (European) — total return per unit staked. Odds 2.50 = €1 bet returns €2.50 (€1.50 profit). Convert: 1 / 2.50 = 40%.
Fractional (UK) — profit relative to stake. 3/2 = €3 profit per €2 staked. Convert: 2 / (3+2) = 40%. To decimal: (3/2) + 1 = 2.50.
American (Moneyline) — positive = profit on $100 stake; negative = stake needed to win $100. +150 means $100 wins $150. Convert: 100 / (150+100) = 40%. Negative: |odds| / (|odds|+100).
Why probabilities sum to more than 100%
Booksum: 105.60% → Overround: 5.60 percentage points
For a standard positive-overround market, the raw probabilities add up to more than 100%. The excess is the overround, often called vig or juice. Margin as a share of booksum is a different ratio: (Q − 1) / Q, where Q is the sum of reciprocal odds. Neither figure is the bookmaker’s observed profit on a particular match. Mixing best prices from different bookmakers, promotional prices or different timestamps can produce a booksum of 100% or less.
The bookmaker margin
For an illustrative complete 1X2 market of 2.10 / 3.50 / 3.40, calculate each reciprocal before rounding:
Booksum Q
105.60%
Overround
5.60pp
Margin / booksum
5.30%
Booksum is Q = Σ(1 / odds). Overround is Q − 1; margin as a share of booksum is (Q − 1) / Q. The latter equals theoretical hold on turnover when stakes are balanced so the payout is the same for every outcome. Actual profit depends on stakes, liabilities and results.
These are worked-example values, not current rates for any named bookmaker. Overround varies by operator, market, competition and timestamp. Compare the complete set of mutually exclusive outcomes from the same market and snapshot; a correct-score market may require an “Any Other Result” outcome to be complete.
Removing the margin: fair odds
Proportional normalization: divide each raw probability by the booksum — the sum of reciprocal odds (approximately 1.056022 in this example), not by the overround itself (approximately 0.056022). This is an explicit margin-allocation assumption, not a way to recover uniquely true probabilities. For additive, power, odds-ratio and Shin alternatives, see the five margin-removal methods compared.
Proportional fair probability: pᵢ = (1 / oddsᵢ) / Σ(1 / odds)
Home = 45.09% | Draw = 27.06% | Away = 27.85% | Total: 100.00%
Fair odds = 1 / fair probability. Home: 2.22 (bookmaker: 2.10). Draw: 3.70 (bookmaker: 3.50). Away: 3.59 (bookmaker: 3.40). In this positive-overround example, the quoted prices are shorter than the proportional fair odds.
The ImpliedScore calculator uses proportional margin removal before fitting a score distribution. Our historical market-calibration study compares de-vigged probabilities with observed outcomes. Results vary by league and probability group; removing margin does not guarantee a fixed accuracy range or a betting edge.
Reading the 1X2 market
| Home / Draw / Away odds | Proportional fair probabilities |
|---|---|
| 1.30 / 5.50 / 8.00 | Home 71.49% · Draw 16.90% · Away 11.62% |
| 1.80 / 3.60 / 4.50 | Home 52.63% · Draw 26.32% · Away 21.05% |
| 2.40 / 3.30 / 3.00 | Home 39.57% · Draw 28.78% · Away 31.65% |
| 4.00 / 3.50 / 1.90 | Home 23.54% · Draw 26.90% · Away 49.56% |
The three prices describe mutually exclusive outcomes under the stated settlement rules, usually the result after 90 minutes plus stoppage time. Remove the margin from the complete 1X2 market before comparing fair probabilities. Draw prices are not restricted to a fixed range. Percentages above are rounded, so a displayed row may differ slightly from 100%.
Reading the Over/Under 2.5 market
Over 2.5 = 3+ total goals. Under 2.5 = 0, 1, or 2 goals. The half-number eliminates pushes.
Proportional fair: Over = 51.95% | Under = 48.05%
| League | Historical Over 2.5 frequency |
|---|---|
| Bundesliga | 59.4% |
| EPL | 54.4% |
| Serie A | 52.1% |
| Ligue 1 | 51.1% |
| La Liga | 48.3% |
The table reports retrospective frequencies from the site study, not probabilities for a new fixture. For a supplied match, the Over/Under market helps constrain total goal expectation, while 1X2 helps constrain the balance between the teams. These signals need not fit one score model perfectly.
Reading the Asian Handicap
| AH line | What it means |
|---|---|
| Home −0.5 | Home must win by 1+ goals |
| Home −1.0 | Win by 2+; exactly 1 = push (refund) |
| Home −1.5 | Home must win by 2+ goals |
| Home +0.5 | Home wins or draws = bet wins |
Asian Handicap prices constrain the goal-difference distribution. A handicap line alone does not determine the expected goal difference, and handicap, 1X2 and totals probabilities are related rather than three statistically independent signals. The main ImpliedScore calculator currently uses 1X2 with optional Over/Under 2.5 odds; it does not fit an Asian Handicap market.
Why understanding odds matters
1. Read the quoted price. Odds of 1.70 encode a raw implied probability of 1 / 1.70 = 58.82%. A single price is insufficient to calculate a margin-free probability: that requires the complete market and a de-vig method.
2. Compare matching bets. Home odds of 2.10 and 2.20 encode raw break-even rates of 47.62% and 45.45%. For the same bet and settlement rules, 2.20 offers a higher return if it wins. The two prices alone do not establish the true probability or how each bookmaker allocated its margin.
3. Read the full distribution. The score calculator removes margin under a stated assumption and fits goal rates to the supplied markets. Its output is a model-implied distribution, not a promise of the final score.
Historical calibration checks whether groups of quoted probabilities line up with observed frequencies. A result in one sample does not establish future accuracy for every league, market or fixture. Neither a lower overround nor a close historical fit guarantees profit.
FAQ
What are closing odds?
The final quoted prices before kickoff. They can incorporate information available later than an opening price, but they are not guaranteed to be correct. The historical ImpliedScore calibration dataset uses Football-Data.co.uk market-average AvgH/AvgD/AvgA columns, not the separate closing-odds columns. See the methodology and data definitions.
Are lower odds always better?
Lower decimal odds encode a higher raw implied probability and offer a lower payout. They are not automatically better value. Assessing value requires comparing the offered price with a defensible estimate of the outcome probability.
Do different bookmakers offer different odds?
Yes. Prices can differ because of information, timing, margin allocation, trading limits and liabilities. Compare the same market, settlement rules and timestamp before interpreting a price difference.
What is a sharp bookmaker?
The term is commonly used for a bookmaker associated with competitive pricing, relatively high limits and prices that respond to informed betting activity. It is not a guarantee that every quote is margin-free or an accurate probability estimate.
Does line shopping create an edge?
For an identical bet, choosing higher odds improves the return if it wins and lowers the break-even win rate. That improvement alone does not establish positive expected value; a reliable estimate of the outcome probability is still needed.