Population Genetics: Wright-Fisher Ensemble (2D)

A single Wright-Fisher run only shows you one possible outcome of a random process. Genetic drift means two populations that start identical can end up completely different — one fixing allele A, another losing it, a third staying balanced. Running dozens side by side turns "fixation probability is roughly 1/(2N)" from an abstract formula into a countable outcome you watch happen across the ensemble.

It is the mean allele frequency across every replicate still contributing data that generation. Under selection alone (drift off) it exactly follows the deterministic Δp equation; with drift on, individual replicates wander away from it while the mean itself stays close to the deterministic prediction as long as enough replicates remain segregating.

Each replicate independently draws 2N alleles binomially from its own post-selection, post-mutation frequency, exactly as the 3D version does for its single population — the only difference here is that this sampling happens once per replicate per generation, so replicates with small N visibly separate from each other far faster than replicates with large N.

That preset gives the heterozygote genotype (Aa) the highest fitness, an overdominance scenario like sickle-cell trait in malaria-endemic regions. Because neither homozygote is favoured, selection pulls every replicate toward the same stable interior equilibrium instead of toward fixation, so the histogram stays clustered near the middle even after many generations, unlike the Dominant Advantage preset where most replicates end up pinned at p=1.