Probability for Betting Odds: The Useful Basics

Probability and complements

A probability lies between 0 and 1. If P(Home win) = 0.45, then P(not Home win) = 1 - 0.45 = 0.55. OpenStax provides the core terminology.

Joint and conditional probability

For events A and B:

P(A and B) = P(A) x P(B given A)

If P(Home win) = 0.50 and P(Over 2.5 given Home win) = 0.60, joint probability is 0.50 x 0.60 = 0.30. Multiplying P(A) by an unconditional P(B) is justified only when the events are independent. OpenStax's independence chapter explains that distinction.

Odds and break-even probability

At decimal odds 2.20, the raw break-even conversion is 1 / 2.20 x 100 = 45.45%. A model estimate of 50% implies fair odds of 1 / 0.50 = 2.00. The difference is model-dependent because the 50% estimate is not observed truth.

Expected value

For decimal odds d and estimated win probability p:

Estimated EV per unit = p x d - 1

At p = 0.50 and d = 2.20, EV is 0.50 x 2.20 - 1 = 0.10 units before costs. At p = 0.40, EV is 0.40 x 2.20 - 1 = -0.12 units. OpenStax defines expected value and variance.

Variance and evaluation

A positive expected-value estimate can lose, and a negative estimate can win. Forecast quality is evaluated across an appropriate set of future observations. Calibration guidance compares predicted probabilities with observed frequencies; it does not prescribe a universal number of bets.

Practical checklist

• State whether probabilities are raw market conversions or model outputs.
• Use conditional probability for related events.
• Include every payoff and cost in expected value.
• Validate on unseen, time-appropriate data.
• Report uncertainty and failed predictions as well as successes.

Use complements as a consistency check

For a binary event, P(not A) = 1 - P(A). If a model assigns 62% to A, it must assign 38% to not A under the same definition. For a three-way football result, the home, draw and away probabilities must sum to 100% before any pricing margin is applied. A total of 96% or 104% signals missing mass, overlapping outcomes or a calculation error in the probability model; OpenStax provides the probability foundation.

Conditional probability is equally important. If P(Home win) is 50% and P(Over 2.5 given Home win) is 60%, the joint probability is 30%. Using an unconditional Over 2.5 probability instead would silently assume independence. OpenStax explains the distinction.

Before using an EV number

• Enumerate every outcome and payoff.
• Check that probabilities are coherent and defined for the same event.
• Include pushes, voids, commission and split settlement where applicable.
• Keep forecast uncertainty visible.
• Evaluate the probability source on observations not used to fit it.

Probability arithmetic can reveal inconsistency, but it cannot create reliable inputs. The quality of the expected-value estimate remains bounded by the quality and timing of the probability estimate.

Related resources

Continue with implied probability or accumulator odds.

Continue learning

• Next guide: Exchange Odds vs Bookmaker Odds
• Related guide: Live Odds vs Pre-Match Odds

Assumptions and limitations

All examples are illustrative and use simplified binary payoffs. Real football markets can have three outcomes, pushes, partial settlement, correlated selections and changing data. Probability mathematics does not supply the probabilities; those require a documented source or model.