The concept of domination for Monte Carlo procedures solving multi-extremal optimization problems is introduced. It is shown that stratified sampling strictly dominates repeated sampling. The organization of stratified sampling is considered.