I have been trying, far less than successfully, to write a program to solve Einstein's Riddle Logic Puzzles, see https://play.google.com/store/apps/details?id=com.rottzgames.logic&hl=en_US&gl=US.​

The program I have written properly executes all of the games ‘rules’. In fact, for a super simple game, it actually came up with the correct answer. However, as I try the more reasonable puzzles, the answer is not correct. For example, in the game Hard:#1, there are five layers and six items. Thus, a proper result would have just six ‘columns’ of results. However, the best I have been able to do is 19 columns, which includes the correct six.

I am clearly missing something. I have looked over this particular game in great detail, but do not see any way to use an indirect approach to help eliminate some of the wrong answers.

So, please, if you have any insights, algorithms, etc., please share. Please.

Is this game based on logic puzzles? If so, you could create a data structure that represents the complete logic grid. Not sure where you would take it from there. How does your program understand the (English) language based clues? Or is your program a help-solve and not a solver?

Also, I just noticed this in AI. Are you using general AI to solve this? If so you should specify. Here, AI is game AI unless specified or obvious.

In answer to your question: I code the board as layers, columns, and items. Each rule is them translated into a layer/item pair, for example a textual rule can become 1C-4B. I then represent each possible solution and work through each rule. I also look for a single item on a layer and recognize that that must be the solution in the column in which it appears.

But, this is not enough! There has to be some type of indirect approach I have yet to discern.