In this work we present optimized noble gas water Leonard-Jones 6-12 pair potentials for each noble gas. Given the significantly different atomic nature of water and the noble gases, the standard Lorentz Berthelot mixing rules produce inaccurate unlike molecular interactions between these two species. Consequently, we find simulated Henry's coefficients deviate significantly from their experimental counterparts for the investigated thermodynamic range (293-353 K at 1 and 10 atm), due to a poor unlike potential well term (epsilon(ij)). Where epsilon(ij) is too high or low, so too is the strength of the resultant noble gas-water interaction. This observed inadequacy in using the Lorentz-Berthelot mixing rules is countered in this work by scaling epsilon(ij) for helium, neon, argon, and krypton by factors of 0.91, 0.8, 1.1, and 1.05, respectively, to reach a much improved agreement with experimental Henry's coefficients. Due to the highly sensitive nature of the xenon epsilon(ij) term, coupled with the reasonable agreement of the initial values, no scaling factor is applied for this noble gas. These resulting optimized pair potentials also accurately predict partitioning within a CO2-H2O binary phase system as well as diffusion coefficients in ambient water. This further supports the quality of these interaction potentials. Consequently, they can now form a well-grounded basis for the future molecular modeling of multiphase geological systems.