Building the initial population.
The software creates an array with each individual identified with its sex and genotype.
Filling the array needs a first set of random choices. Genotype and sex are chosen according to the user's settings for the population.
For instance if the current settings are: two alleles sharing a 50% frequency, it's only a 0.5 probability that have been chosen. The final result can (and most probably will) give an actual frequency slightly different.
The larger the population, the thinner the gap.
The idea is that out of a former very large population, was extracted a smaller, randomly chosen, sample. The resulting gap between the mother and child population is a good simulation of the founder effect.