Each bee carries three independent breeding values in [0,1]: temperament (gentleness), body size (a proxy for productivity) and cuticle colour depth. Selecting a criterion and breeding a generation keeps only the top fraction of the colony by that trait as parents; every offspring's value is the average of two randomly-paired parents plus Gaussian mutation noise, clamped back into range.
child = clamp( (parentA + parentB)/2 + N(0, sigma) )
h² = V_A / V_P (additive genetic variance / total phenotypic variance)
- Selection intensity — the top % of the colony (by the chosen trait) kept as breeders; a smaller number narrows the gene pool faster.
- Mutation strength — random noise (sigma) added to each offspring's trait; it is the only source of new variation once the founder population's diversity is used up.
- Genetic diversity — the variance of the selected trait across the live colony; when it collapses toward zero, further selection barely moves the population and inbreeding risk rises.
Traits with high heritability (h² close to 1, like colour) respond fast and predictably to selection. Traits with low heritability (h² close to 0, dominated by environment) barely shift even under intense, sustained selection — this is a simplified single-locus-style analogue of that real breeding dynamic, not a full quantitative-genetics model.