Coarse-grained (CG) simulations are widely used for observing molecular phenomena, such as self-assembly in biological systems. CG models are typically developed by optimizing the model's potentials to match properties obtained from atomistic simulations. Typically in simulations of self-assembly an explicit water solvent is used; however, for coarse-graining methods that require an explicit mapping of atomistic properties to the CG level for potential optimization, water is usually mapped with one bead representing one water molecule, forcing the solvent to be the most computationally expensive component of the simulation. Although other methods where a bead in the CG model represents multiple water molecules are more computationally efficient, the derivation of their potentials are not systematic and may not be robust. To address this problem we have devised a novel strategy to coarse-grain water to a multiple molecule per bead model. Results will be presented for several CG water models solvating hydrophobic and hydrophilic CG molecules to demonstrate what degree of detail (with respect to water) is necessary to mimic behavior seen in atomistic simulations.