Drift is just luck, repeated
Two identical alleles, no advantage for either. Watch a finite population forget its variety by chance alone — and see why fewer copies means a wilder walk.
Chance, not selection
Both alleles are neutral: a bead of either colour is equally likely to be copied. Each generation, every offspring picks a random parent, so the number of A beads is Binomial(N, p). Its expected value is the old count, but the actual count wanders.
Why small populations drift faster
The spread of one draw is √(p(1−p)/N). Ten beads can easily lurch by a few; two hundred barely move. Try N = 6 against N = 200 with the same start: the small ones hit the walls within a few dozen generations.
Walls that never let go
At frequency 0 (A lost) or 1 (A fixed) there is nothing left to sample from, so the draw returns the same frequency forever. Drift removes variation, and a neutral allele fixes with probability equal to its starting frequency.
Simplified model: haploid Wright–Fisher, N gene copies, non-overlapping generations, no mutation, migration or selection. Diploid organisms have 2 copies per individual.