Thursday, May 6, 2021

Puzzle #350 - Making a Puzzle is as (Easy as ABC)

So, 350 puzzles. That's a big milestone. Maybe I'd have today's puzzle be for hyping up my upcoming puzzle contest, but no, I'm planning on posting about that once the final instructions booklet is ready. Maybe today would have another giant puzzle, or some crazy hybrid rules.
Or maybe, today I'll take a moment to talk about the construction process of a puzzle like this, and peel back the curtain. Not every puzzle comes out exactly the way I want it to and I often have to backtrack, and this puzzle was no exception.
The inspiration for making this puzzle was simple- on the thinky-puzzles Discord, skymoo wondered how to create "ABC logic puzzles", and suggested a constraint solver to generate solved boards, and got stuck on ensuring the puzzle could be solved from the placed clues. Again, the proposed solution was to write a solver, but that moved the problem to which clues are kept and how to get a desired difficulty. I responded quickly with a brief overview of how I made Puzzle #292 and outlined the general construction process- which is solving a puzzle as you make it, placing clues to give answer progress and then using that answer progress to guide the next clues.
I presented these 4 clues as a possible simpler start, knowing that they immediately placed two letters and two pairs. I then turned this start into a full puzzle- I wanted to try for symmetric clues, and the obvious continuation was to repeat the same deduction in the other corner.
In order to get the same pattern of empty cells, the clues on the extremes would have to be the same. That resulted in this board state, regardless of what clues I chose- then I just tried A, B and C as the duplicate clue.













I quickly found that A and B would both break the same way (note the A farthest from the A in the first grid- due to the pseudo-symmetry, B would do the same thing), and that C would specify the corners but leave 3 solutions. Worse, there would be no way to specify the remaining cells and it would be the exact same deduction in both corners- not ideal!

So I then tried a slight variation on the same idea, placing ABC down the right side in order to force the rightmost column to read BAC--. That actually resolved the entire puzzle on its own! I still placed the remaining clue (a B) as it didn't really affect the crucial steps of the puzzle, I was aiming for an easy demonstration puzzle, and I liked having the clue symmetry.

Should I do another post like this?

No comments:

Post a Comment