Possessing control over the molecular size (molecular weight/chain length/degree of polymerization) distribution of a polymeric material is extremely important in applications. This is manifested de facto by the extensive contemporary scientific literature on processes for controlling this distribution experimentally. Yet, the literature on computational techniques for achieving prescribed molecular size distributions in simulations and exploring their impact on properties is much less abundant than its experimental/technical counterpart. Here, we develop—on the basis of united atom melt simulations employing connectivity-altering Monte Carlo moves—a new Metropolis selection criterion that drives the multichain system to a prescribed but otherwise arbitrary distribution of molecular sizes. The new formulation is a generalization of that originally proposed [P. V. K. Pant and D. N. Theodorou, Macromolecules 28, 7224 (1995)], but simpler and more computationally efficient. It requires knowledge solely of the target distribution, which need not be normalized. We have implemented the new formulation on long-chain linear polyethylene melts, obtaining excellent results. The target molecular size distribution can be provided in tabulated form, allowing absolute freedom as to the types of chain size profiles that can be simulated. Distributions for which equilibration has been achieved here for linear polyethylene include a truncated most probable, a truncated Schulz–Zimm, an arbitrary one defined in tabulated form, a broad truncated Gaussian, and a bimodal Gaussian. The last two are comparable to those encountered in industrial applications. The impact of the molecular size distribution on the properties of the simulated melts, such as density, chain dimensions, and mixing thermodynamics, is explored.