Hey /sci/, I hope this kinda thread is allowed here. I'm more often lurking the discussions here.
I don't really game much, but when I play I do like puzzle games. Some of them are really hard like snakebird or stephen's sausage roll or spachechem. One mistake can mean that you won't be able to solve the puzzle.
And this got me wondering are there any mathematical methods/rules one could apply to such or to puzzles in general? I tried searching for stuff but I did not find anything, or is it impossible because of the always different rulesets? (even though many aspects are often simmiliar)
Hope the question makes sense, english is not my naitive language.
>>8600556
Brute force or machine learning.
If you want to create formulas that are more condensed, then that would be dependent on the ruleset.
>>8600556
Intuition from experience.
Logical thinking.
Brute force.
>>8600561
>>8600570
Ah that's what I thought the answers would be, been thinking about this for some time.
Had some ideas about min-maxing, but that doesen't work.
And I thought maybe some graph-theory or topology could be applied, but I do not know enough about that stuff.
Backtracing is the only thing that somewhat works I think, but again depending on the game.
>>8600585
>>8600585
>Had some ideas about min-maxing, but that doesen't work.
That could work. That's like machine learning or as the other poster put it "intuition from experience".
The robot or the human could learn how to reduce the amount incorrect moves. Probably, the human/robot would not actually make the move, but would mentally make the move to see what would happen. So the goal would really be to reduce the necessary mental moves (i.e. trial/error). That's a minimization task that an AI or human could use to learn.
Probably the robot or the human would break the game into sub-games in order to minimize the amount of trial/error necessary to find the correct path (kind of like how that go AI loosely works).
tree search.
>>8600556
>spacechem
>hard
>>8600556
There is a bunch of mathematics on how to solve puzzles AFTER you have precisely formalized the rules. I'm not aware of anything substantial on the topic of going from a puzzle that a human understands to a mathematical model.