Relative frequency of a random event
The classical definition needs equally likely outcomes, which are not always known in practice. So experiments are carried out: if event A happens in M of N trials, M is the frequency of A and W(A) = M : N is its relative frequency. As the number of trials grows, the relative frequency settles around some number, which is taken as the statistical probability of the event. For example, in the coin experiments of Buffon and Pearson the relative frequency of heads came out very close to 0.5. Jacob Bernoulli justified this as the law of large numbers: for many trials W(A) ≈ P(A). With few trials, the relative frequency can differ noticeably from the probability.
A pupil who recorded rain on 6 of the 30 days of April finds the relative frequency of a rainy day W = 6 : 30 = 0.2. They can say “it rained about one day in five”, but that is no guarantee for next month.
A meteorologist computes relative frequencies from many years of records, a quality inspector finds the share of faulty products, and a public-health researcher estimates how common something is.
Take the first 40 letters on a page of a book, count how many times the letter “e” appears, and find W = m : 40.
“Coin series”: each pair tosses a coin 20 times and records the heads; then the whole class’s results are added, W is computed and compared with 0.5.