Mathematics
The Geometric Distribution: Waiting for the First Success
Quick fact
Even if the probability of success is high (e.g., 0.8), the expected number of trials until the first success is only 1/p, but the probability of needing more than the expected value is surprisingly large (about 37%).
Why this is interesting
Have you ever wondered how many tries you'll need before you finally roll a six? The geometric distribution gives a precise answer, and it often defies intuition.
Read the full explanation
Understanding The Geometric Distribution: Waiting for the First Success
Think of flipping a coin until you get heads. Each flip is a Bernoulli trial with two outcomes (success/failure). The number of flips needed to see the first heads is a random variable. If the coin is fair (p = 0.5), you might get heads immediately or wait several flips. The geometric distribution describes the probabilities for each possible number of trials. For any single trial, the chance of success is p, and each trial is independent. So the probability that the first success occurs on the k-th trial is the probability of having (k-1) failures followed by one success: (1-p)^(k-1) p. This is the probability mass function (PMF). For a fair coin, the most likely outcome is 1 (probability 0.5), but the distribution has a long tail: you could wait many flips.
A deeper explanation
The geometric distribution emerges from the assumption that each trial is independent and identical. This creates a 'memoryless' property: if you haven't succeeded after k trials, the probability of needing t more trials is the same as the probability of needing t trials from the start. This is because the failures you've already seen don't change the probabilities of future trials. Mathematically, this is expressed as P(X k+t | X k) = P(X t). This property is unique to the geometric (and exponential) distributions. The expected number of trials until the first success is 1/p. For example, if p = 0.1, you'd expect 10 trials on average. However, the variance is (1-p)/p^2, which is large when p is small, meaning waiting times can be highly variable. The geometric distribution is the discrete analogue of the exponential distribution and is used in reliability theory, where it models the number of operations before a device fails.