I think you'd have to reduce the problem set/space for that to be effective (which is a very similar concept to reducing how you define your state space in AI, if you're familiar with that). If the problem set has too many variables, I think it becomes too challenging to analyze.

My approach is to rely very heavily on heuristics in order to screen potential treatments. I wrote some steps describing how to do that elsewhere in this thread.

Yeah, I guess I really don't have a handle on the complexity.

Your heuristics seem useful. I've been just trying stuff for a month and seeing how they go. Trusting the professionals to guide me in semi-useful directions.

Thanks.