Efficient Monte Carlo sampling for long molecular chains using local moves, tested on a solvated lipid bilayer Abstract: An improved acceptance criterion for local move Monte Carlo method (in which trial steps change only seven consecutive torsion angles of a polymer) is introduced. In contrast to previous methods, not all possible loop closures are considered, but only the most structurally conservative one. By filtering such moves with the `reverse proximity criterion' introduced here, detailed balance is satisfied. Furthermore, the new method is ergodic, and is shown to be significantly more efficient than previous methods when applied to a fully solvated hydrocarbon chain with bulky sidechains as well as a fully solvated lipid bilayer.