Sampling and Experiments · Superchallenge problem

Four pairs of identical twins (8 people) volunteer for an experiment. Instead of using the twins as blocks, the researcher splits the 8 people completely at random into a treatment group of 4 and a control group of 4. What is the probability that, by chance, every pair of twins is split (one twin in each group), as a matched-pairs design would guarantee?