A binomial random variable counts successes in a fixed number of independent trials with the same success probability. Recognizing those conditions is the first task; applying the formula to an unsuitable process produces precise-looking but incorrect probabilities.
Check the four binomial conditions
There must be a fixed number n of trials, two outcome categories per trial, independent trials, and a constant success probability p. “Success” is simply the category being counted, not necessarily a desirable outcome.
If sampling without replacement from a finite population, trials are not exactly independent. The binomial model can be a reasonable approximation when the sample is no more than about 10% of the population and the success probability changes negligibly. State that approximation when it matters.
A process that continues until the first success does not have a fixed trial count and is not binomial. A fixed number of trials with changing success probabilities also fails the standard model.
Why the probability formula has two parts
If then
The factor gives the probability of one particular arrangement with successes. The binomial coefficient counts how many arrangements have that same success count. Omitting it counts only one order of outcomes.
Example: exactly two successes
A player independently makes each of five shots with probability 0.4. The chance of exactly two made shots is
An arrangement such as success-success-failure-failure-failure has probability 0.03456, but there are ten possible placements of the two successes. The binomial coefficient accounts for all of them without listing them individually.
The probabilities for k from 0 through n sum to 1. A histogram of those probabilities displays the distribution, with bars at integer counts rather than a continuous curve.
Binomial probabilities for five independent trials with success probability 0.4: counts zero through five have probabilities 0.07776, 0.2592, 0.3456, 0.2304, 0.0768 and 0.01024.
Exactly two successes is the tallest bar, but other counts remain possible. The bar heights together sum to 1.
Describe this figure
Binomial probabilities for five independent trials with success probability 0.4: counts zero through five have probabilities 0.07776, 0.2592, 0.3456, 0.2304, 0.0768 and 0.01024.
0: 0.0778; 1: 0.2592; 2: 0.3456; 3: 0.2304; 4: 0.0768; 5: 0.0102.
Example: translate “at least” using a complement
For the same player, the probability of at least one success is This is simpler than summing the probabilities of one through five successes.
“At least two” means so subtract both zero and one success: “More than two” means a different event. Write the inequality before selecting a calculator command.
A cumulative probability includes . To find use not The one-unit adjustment is a frequent source of errors.
Mean and spread come from adding indicators
For a binomial count, and Each trial contributes an indicator equal to 1 for success and 0 otherwise. Adding expected values gives np; independence lets variances add, giving
For five shots with equals 0.4, the mean is 2 and the standard deviation is shots. The mean need not be a possible integer outcome in every binomial model; it is a long-run average across repeated sets of trials.
Shape and normal approximation
The distribution is roughly symmetric near and skewed when is near zero or one, especially for small . A normal approximation becomes reasonable when both and are sufficiently large, commonly at least 10.
When approximating discrete counts with a continuous normal curve, a continuity correction can improve the match: for example, corresponds approximately to a normal value below An exact binomial calculation is preferable when available and appropriate.
Try it yourself
A component fails independently with probability 0.1. Among 8 components, find the probability of no failures and the expected number of failures.
Answer: and An expected count below one does not mean failure is impossible; it summarizes repeated groups of eight components.
What to remember
Verify the model conditions, distinguish exact and cumulative events, and use complements carefully. The count's mean is np and its variance is