Has there been any work done on understanding what god's solution is doing such that a human might be able to replicate it?

What I am imagining is analyzing a large database of god's solutions to see if one can say simple things like "each turn tries to maximize the number of corner/edge cubie matches" or something like that. IN other words, is there some "gradient descent"-like behavior one can glean by studying, computationally, a large number of god's solutions in action.

I'm a cube newbie but trained in mathematical optimization, hence the "gradient descent on combinatorial graphs" way of thinking about this problem. Any pointers/directions to ideas about this would be greatly appreciated.