Betting Odds vs Probability: How to Compare Them

Bookmakers publish odds. GoalStatsHub publishes probabilities. They describe the same thing, how likely an outcome is, but odds carry the bookmaker’s margin and probabilities don’t. This guide shows how to convert one into the other, how to strip the margin out, and why a small gap between the two numbers is not the opportunity it looks like.

From odds to implied probability

For decimal odds, the format used across most of Europe, the implied probability is one divided by the odds:

• odds of 2.00 imply 1 ÷ 2.00 = 50%;
• odds of 2.50 imply 40%;
• odds of 1.25 imply 80%.

The other formats convert to decimal first:

• Fractional (UK): add one. 6/4 becomes 1 + 6 ÷ 4 = 2.50, so 40%.
• American, positive: divide by 100 and add one. +150 becomes 2.50.
• American, negative: divide 100 by the number and add one. −200 becomes 1.50, so 66.7%.

Going the other way, the “fair” odds for a probability are one divided by it. Our model’s 46.4% for a home win corresponds to fair odds of about 2.16.

The margin: why the percentages add up to more than 100%

Suppose a bookmaker prices a match like this:

The implied probabilities add up to 104.8%. The extra 4.8% is the bookmaker’s margin, often called the overround or vig. It is how the bookmaker earns money regardless of the result. To compare odds with a model fairly, divide each implied probability by the total (here 1.048). That gives the last column, the bookmaker’s view without its margin.

Margins vary a lot. Main markets in top leagues are usually priced tightly; lower leagues, exotic markets such as correct score, and accumulators carry much more.

Comparing a model with the market

Say our model gives the same match 46.4% home, 25.8% draw and 27.8% away, the figures from the worked example in how prediction models work. Against the margin-free market numbers, the model is one point higher on the home win, two points lower on the draw and one point higher on the away win. That is close agreement, which is normal: bookmakers and statistical models look at much the same information.

The tempting idea is to back whatever the model rates higher than the market. The expected return of a bet is the probability times the odds, minus one:

• home win: 0.464 × 2.10 − 1 = −2.6%
• draw: 0.258 × 3.40 − 1 = −12.3%
• away win: 0.278 × 3.60 − 1 = +0.1%

Even where the model is a little more optimistic than the market, the margin eats the difference. And this calculation assumes the model’s figure is the true probability. It isn’t. Every model has error bars, and a gap of one or two percentage points is well inside them.

Why the market is hard to beat

Odds move as money comes in and as news breaks: team sheets, injuries, weather. By kick-off the price on a major match reflects a great deal of information that a results-based model never sees. A model can still be useful, as a sanity check on a price, as a way to see how markets relate to each other, or for less-watched leagues, but a percentage on a website is not proof that a bet has value.

Probabilities and losing runs

A probability describes many matches, not one. A 70% favourite is expected to fail roughly three times in ten. Ten independent 70% picks all landing happens only about 3% of the time, so streaks of misses are normal even for a well-calibrated model. Our guide to judging predictions explains how to tell bad luck from a bad model.

Using our numbers sensibly

• Convert any odds you see to a probability before comparing them with ours.
• Remove the margin first, or every market will look overpriced.
• Treat small differences as noise, not as a signal.
• Check the confidence dots: a one-dot prediction rests on little data.

18+. Our percentages are information, not betting advice. If you bet, set a limit you can afford to lose and read our responsible gambling page.