Using the Complement Rule in Probability

Some probability questions are easier to solve by counting what does not happen. The complement rule formalizes that shortcut. If event A either happens or does not happen, those two possibilities cover every outcome and do not overlap. Their probabilities must add to 1. So the probability that A occurs is 1 minus the probability that A fails to occur.

Define the event before using the rule

The complement of A is often written Aᶜ or ‘not A.’ If A is “a randomly selected student takes the bus,” then Aᶜ is “the student does not take the bus.” If A is “at least one six appears,” then Aᶜ is “no six appears.” The second event is a single clear alternative even though A includes several possible outcomes: exactly one six, exactly two sixes, and so on.

This definition matters. The complement of “more than 3” is “3 or less,” including 3. It is not “less than 3.” The complement of “at least 1” is “0,” while the complement of “at most 1” is “2 or more.” Translate boundary language precisely before calculating.

The complement identity

Since A and Aᶜ are disjoint and together include all outcomes, P(A) + P(Aᶜ) = 1. Rearranging gives P(A) = 1 − P(Aᶜ), and likewise P(Aᶜ) = 1 − P(A). If the probability that a component fails during a given test is 0.08, and success and failure are the only outcomes, then the chance it does not fail is 1 − 0.08 = 0.92.

The complement rule works with a theoretical model or an empirical probability, as long as the event and its complement exhaust the possible outcomes. If a question includes a third possibility—such as a test result being positive, negative, or inconclusive—then ‘not positive’ includes both negative and inconclusive. Do not silently equate a complement with only one of several alternatives.

Why ‘at least one’ often points to a complement

Suppose a fair coin is flipped five times. To find the chance of at least one head, you could add the probabilities of getting exactly one, exactly two, three, four, or five heads. That is possible, but there are many cases. The complement is simpler: the only way to get no heads is for all five flips to be tails. The probability of that is (1/2)⁵ = 1/32. Therefore the probability of at least one head is 1 − 1/32 = 31/32.

For n independent trials, if the probability of success on each trial is p, the chance of failure on one trial is 1 − p. The chance of no successes in all n trials is (1 − p)ⁿ, because independence lets us multiply the failure probabilities. Thus the probability of at least one success is 1 − (1 − p)ⁿ. This formula is not valid without the stated assumptions: the trials need to be independent and have the same success probability for this exact expression.

Worked examples

At least one defective item

A process independently produces a defective item with probability 0.03. For a batch of four items, what is the probability at least one is defective? The probability one item is not defective is 0.97. The probability all four are not defective is 0.97⁴ = 0.88529281. Subtract from 1: 1 − 0.88529281 = 0.11470719, or about 11.47%. Adding four times 3% would give 12%, which overcounts cases with multiple defective items.

At least one face card

Draw two cards from a standard deck without replacement. The probability of at least one face card (jack, queen, or king) is one minus the probability that neither card is a face card. There are 40 non-face cards. The probability the first card is a non-face card is 40/52; after that, 39 non-face cards remain out of 51, so the probability the second is also a non-face card is 39/51. The complement is 1 − (40/52)(39/51), approximately 0.4118, or 41.18%. Because the draw is without replacement, the two factors are conditional rather than identical independent chances.

Exactly one versus at least one

A common error is to use a complement for the wrong event. “At least one success” is the complement of zero successes. “Exactly one success” is not: the complement of exactly one includes zero successes and two or more successes. To find exactly one success in several trials, count or model the one-success cases directly unless an easier partition is given.

Use a complement with equally likely outcomes

For a finite sample space with equally likely outcomes, probability can be found by counting favorable outcomes over total outcomes. Sometimes the unfavorable set is shorter. If a password code is chosen uniformly from all three-digit strings from 000 through 999, the probability of at least one 7 can be found as 1 minus the probability of no 7. Each position then has 9 allowed digits, so the no-7 count is 9³ out of 10³; the desired probability is 1 − 729/1,000 = 271/1,000. This assumes each digit is chosen independently and repetition is allowed.

A reliable solving sequence

1. Write the event the question asks for, especially words such as at least, at most, more than, and exactly.
2. State its complement in plain language, including boundary values correctly.
3. Decide whether the complement is easier to count or calculate.
4. Use the sample space, conditional probabilities, or independence rule that actually fits the process.
5. Subtract the complement probability from 1.
6. Check that the result is between 0 and 1 and that the complement plus the answer equals 1.

Common mistakes

• Subtracting from 100 instead of from 1 when probabilities are written as decimals. With percentages, subtract from 100%.
• Using (1 − p)ⁿ when trials are dependent or have different success probabilities. Multiply the correct sequential conditional probabilities instead.
• Misreading at least or at most and excluding an endpoint that belongs in the event.
• Treating the complement of one category as a single other category when the sample space has several remaining outcomes.
• Adding the chance of success across repeated trials without accounting for overlap among ways success may happen more than once.

Exam takeaway

When a prompt says “at least one,” check whether “none” is easier to calculate. Define the complement precisely, find its probability using the actual dependence structure, then subtract from 1. The rule itself is simple; most errors come from a badly defined opposite event or an unjustified independence assumption.