Positive Expected Value Betting: Formula and Examples

Short answer: for decimal odds, expected return per unit staked is your probability × decimal odds - 1. A positive result is a modelled edge, not guaranteed profit. The conclusion is only as good as the probability estimate, price and assumptions behind it.

Positive EV formula

For a £1 stake at decimal odds O and estimated win probability P:

Expected return = P × O - 1

Example: you estimate a 48% chance and can take odds of 2.20.

0.48 × 2.20 - 1 = 0.056, or +5.6% expected return per unit staked.

If your true chance were only 43%, the same bet would have expected return of 0.43 × 2.20 - 1 = -5.4%. A five-percentage-point probability error reverses the decision.

Break-even probability

The break-even probability is 1 divided by decimal odds. At 2.20 it is 45.45%. Your probability must be higher than that threshold after allowing for uncertainty and costs.

Use our odds converter to check common formats.

Why the market probabilities exceed 100%

Convert every outcome in a market with 1 divided by its odds. The total normally exceeds 100%. That excess is the overround, a simple view of the bookmaker margin before accounting for how it is distributed across outcomes.

For a two-outcome market priced at 1.91 and 1.91, each side implies 52.36%, for a total of 104.71%. Treating either 52.36% figure as a fair forecast would include the margin. Use the vig calculator to normalize a complete market.

Research on bookmaker pricing shows that margin is not always spread evenly, particularly between favourites and longshots. A recent Oxford Economic Papers study discusses how market structure can contribute to favourite-longshot bias.

Where your probability should come from

A defensible estimate needs a repeatable method. Examples include a calibrated statistical model, a market-based estimate with the margin removed or a documented combination of independent forecasts. “The team looks strong” is not a probability model.

Record:

• the data available before the bet;
• the model version and assumptions;
• the probability before seeing the result;
• the offered and accepted odds;
• all fees, commission and rejected stakes;
• the event result and closing reference price.

Calibration matters more than confidence

If selections assigned 60% probability win about 60% of the time over a large relevant sample, that probability band is calibrated. Accuracy alone is not enough because always choosing the favourite can look accurate while still producing poor prices.

Use out-of-sample data. A model tuned and judged on the same historical matches can learn noise. Avoid changing the method after every loss or reporting only the sports and periods where it happened to work.

Expected value is not the same as a likely win

A bet can be positive EV and more likely to lose than win. At fair odds of 4.00, a 27% estimate produces +8% expected return even though the outcome is expected to lose 73% of the time. Variance will dominate small samples.

The reverse is also true. A 75% favourite at odds of 1.25 has expected return of -6.25%. Being likely to win does not make the price valuable.

How to test a claimed edge

1. Freeze the method before the evaluation period.
2. Log every qualifying selection, including those you could not place.
3. Use the price actually accepted, not the highest historical quote.
4. Separate model performance from execution failures.
5. Check calibration by probability band.
6. Compare with a consistent closing-price reference.
7. Report sample size, turnover, return and uncertainty.

Published betting studies can find inefficiencies in a particular dataset and period, but that does not make the result universal. Market prices often absorb public information quickly, as shown in research on football price reactions after goals.

Stake sizing

A positive estimate does not justify risking the whole bankroll. Probability error, correlation and limits can make several apparent edges less independent than they look. Use a small, pre-set bankroll fraction and reduce stakes when the estimate is uncertain. Never increase a stake to recover a loss.

Common positive-EV errors

• Using implied probability without removing margin.
• Confusing a likely outcome with a valuable price.
• Building and testing a model on the same data.
• Ignoring odds movement between selection and acceptance.
• Excluding rejected or unavailable bets from the record.
• Calling a short winning run proof of an edge.
• Ignoring commission, currency and withdrawal costs.

Responsible gambling

Expected value describes an average under uncertain assumptions. It cannot make short-term losses safe or affordable. Use only money set aside for entertainment, set loss and time limits and stop if betting becomes difficult to control.

Positive EV FAQ

What does +EV mean in betting?

It means your estimated probability and the offered odds produce an expected return above zero. It remains an estimate, not a guaranteed result.

How do I calculate EV from decimal odds?

Multiply your win probability by the decimal odds, then subtract 1. Multiply by 100 to express the result as a percentage.

Can a losing bet have been positive EV?

Yes. Expected value applies across repeated comparable decisions. Any individual positive-EV bet can lose.