Survival probabilities are not straightforward toobtain when observation periods of individuals differ in length. The Kaplan–Meier theatre is a classroom activity, which starts by a data collection exercise where students imagine sailing on the Titanic. Several students ‘fall in the water’ where they are observed by a neighbouring student while they try to hold their breath as long as they can. The observation periods are designed such that some students ‘drown’ and other ‘survive’ until the end of the experiment. Based on the data collected, it is explained why even simple statistics may fail when applied naively. For example, the frequency of students who ‘survived’ 40 s would generally be an estimate of the probability to survive 40 s. However, an issue occurs when there is a student who ‘survived’ but was observed only for 35 s. Then, it is unknown (censored) if the student ‘drowned’ between 35 and 40 s. The Kaplan–Meier method assumes that censored individuals have the same survival chances as the individuals who are still observed. During the Kaplan–Meier theatre, students perform a clever algorithm (Efron 1967), which translates the assumption into action and results in the Kaplan–Meier estimate of survival.