Conditional Probability: Concepts, Formulas and Examples

The formula is P(B | A) = P(A ∩ B) / P(A), provided P(A) is greater than 0. Knowing that A happened shrinks the world: outcomes outside A become impossible, and B is re-evaluated inside the smaller universe of A.

This idea matters because real information always arrives conditionally. A medical test result, a witness statement, or an early election trend each update a probability by restricting the sample space to what is now known. Conditional probability is the formal machinery behind that updating, and it is the concept from which the multiplication theorem, total probability and Bayes theorem all grow. Without it, probability could only describe situations with zero prior knowledge.

Key Concepts

• Notation P(B | A): the vertical bar means given; A is the conditioning event, B the event being evaluated.
• Reduced sample space: given A, the effective sample space becomes A itself, and only the part of B inside A counts.
• Requirement P(A) > 0: you cannot condition on an impossible event, because division by zero is undefined.
• P(B | A) vs P(A | B): different numbers in general; the probability a card is a king given it is a face card is not the probability it is a face card given it is a king.
• Independence link: A and B are independent exactly when P(B | A) = P(B); the information in A is useless for predicting B.

Formulas

Definition: P(B | A) = P(A ∩ B) / P(A), valid when P(A) > 0. Rearranged, the multiplication rule: P(A ∩ B) = P(A) × P(B | A). Counting shortcut for equally likely outcomes: P(B | A) = (outcomes in A ∩ B) / (outcomes in A). No fairness assumption is needed for the general formula.

Worked Example

One card is drawn from a 52-card deck, and you are told it is a face card (jack, queen or king). What is the probability it is a king? Let A be the card is a face card and B it is a king. P(A) = 12/52 and P(A ∩ B) = 4/52, so P(B | A) = (4/52) / (12/52) = 4/12 = 1/3. The counting shortcut agrees: among the 12 face cards, 4 are kings, giving 4/12 = 1/3. Knowing the card is a face card raised the probability from 4/52, about 0.077, to 1/3, about 0.333.

Common Misconceptions

• P(B | A) equals P(A | B): Almost never. The probability of a positive test given disease differs wildly from the probability of disease given a positive test; confusing them is the classic prosecutor's fallacy.
• Conditioning always raises the probability: It can lower it too; information working against B reduces P(B | A) below P(B).
• You can condition on anything: Only on events with positive probability. Conditioning on a zero-probability event is undefined.

Real-World Applications

Spam filters compute P(spam | word pattern) from training data. Insurers price policies using P(claim | driver profile), and every recommendation engine ranks items by probabilities like P(click | user history).

Conditional probability P(B | A) = P(A ∩ B) / P(A) re-evaluates B inside the shrunken universe where A is known to have occurred. It formalises how new information updates chances, distinguishes P(B | A) from P(A | B), and characterises independence as P(B | A) = P(B). It is the building block of the multiplication theorem, total probability and Bayes theorem.