Time for some help.
I originally stated that I hope to harvest the emergent wisdom of the gamer community. I’m finally ready to start that. So far, the AI has been more artificial than intelligent.
Can anyone tell me under what conditions c-g bonds are required to close loops? I would like to come up with a general rule that can be applied to many different puzzles.
A Little Fuzzy Non-logic (apologies to Mozart).
I’m still milking puzzle 11 for all I can get out of it. I has some interesting properties and it only takes 11 minutes to run my program on it (the next puzzle takes a full half hour). I know that c-g bonds are useful in closing loops. P11 has five loop closures that could use c-g bonds. However, a condition of the puzzle is that there are no more than 2 c-g pairs.
If I specify that all loops are closed by c-g bonds and there are a maximum of 2 in the puzzle, that produces a contradiction and exactly zero solutions are found. If I remove the conditions that c-g closures are used, then I get 186,624 results, very few of which are actually solutions. If I specify a 30% probability that a c-g bond is required, about 40 solutions are produced (the results are not constant, now that probability is involved), and some of those are correct. I call this non-logic because it is a scatter gun approach. Even if totally deterministic logic is not practical, some ideas like “c-g bonds are more likely to be needed in this situation than that one”, would be very helpful.
So, who has some ideas?