Optimization Algorithm Setup Question

I am trying to come up with an algorithm to optimize the placement of objects in a grid (e.g. 5 x 7). Each grid space (element) can have 6 discrete states. I have some statistical data based on random grid states and would like to come with an algorithm to come up with "optimized" grid patterns based on the statistical data. For example, if I have the following obtained data (GridPatternXYZ: Statistical Value) from experiments:

GridPattern001: 0.35 GridPattern002: 0.22 GridPattern003: 0.71 GridPattern004: 0.33 .. GridPattern020: 0.63

I would like to come up with an algorithm to generate a GridPattern that will most likely obtain a value closer to 1.0 in an experiment.

Has anyone come across such a problem? Any advice on where to start?