Monday 30 May 2022

pzprPT Masyu experiment

I thought I'd try the same thing as the previous post using the pzprRT interface - this time with Masyu.  

UPDATE: I managed to put out 15 puzzles, which seems absolutely insane to me.  I had speculated that the interface might be a bit more fiddly with Masyu, because you can't just type in numbers - however what won through is the fact that you need far fewer clues to get a valid Masyu puzzle.  So there we go, churn out a puzzle in an average of 2 minutes.  Make of that what you will!

Solved in 0'50.  A fairly tricky puzzle in the end, although enabled with a repetition of my favourite diagonal pairs in the corner.

Solved in 0'44.  Mostly smooth, although a couple of tricky deductions.  My initial impressions are this has been a bit more successful than the Slitherlink experiment to come up with puzzles that don't feel too artificial.

Solved in 1'41.  This is the first one I feel like I missed something - maybe too early to say if that points at a computer generated hallmark, although I didn't really like this as much.

Solved in 0'33.  The theme suffered in order to get something unique out, but this was a lovely smooth puzzle to solve.  I might try and rescue this at a later date!

Solved in 1'09.  Solved nicely enough, but with a couple of moments to sit and think.  But these aren't feeling so artificial to me.  I wonder if that's because I'm a reasonably proficient Masyu solver and just get these things quicker than I do with Slitherlink?  Or indeed if there's an upper limit as to how hard you can make Masyu?

Solved in 0'26.  Finally got an anti-symmetric one going, and it's just as good as any other in this style that I've ever written.  Interesting!

Solved in 0'27.  On a roll with anti-symmetry.  This one has a really nice design actually - maybe it compares favourably to most of the others in this style I've ever written.  Hmmm!

Solved in 0'49.  I'm beginning to enjoy these!  This one had a couple of my favourite patterns to get things going.

Solved in 0'36.  Fine again!  A perfectly reasonable puzzle with a don't close the loop early theme.

Solved in 0'35.  Felt like a perfectly good themed puzzle that solved very smoothly to me.

Solved in 1'24.  Admittedly I guessed here - I wasn't expecting a puzzle with a random selection of white pearls to give anything other than something that needed guesswork.

Solved in 0'30.  I guess all-black pearl puzzles are always going to be less interesting.

Solved in 0'52.  This attempt at anti-symmetry didn't quite work out, but a perfectly decent puzzle!

Solved in 0'37.  Another nice solve!

Solved in 0'36 - more antisymmetry and solved well enough.

Sunday 29 May 2022

pzprRT slitherlink experiment

Perhaps I'll reflect more on this after I finish the experiment - but for the time being I'm going to live-blog a slitherlink puzzle setting process which relies entirely on the hugely impressive pzprRT project.  Let's see how many puzzles I can publish in the next 30 minutes...

UPDATE: The answer is 9 puzzles, which feels like a lot for 30 minutes worth of work.  I have now gone through and solved them all - they are unique and cover a surprising range of difficulty levels.  Some of them are more fun to solve than others!

Here are the puzzles:

Solved in 1'02.  The solve was a bit disjointed solving symmetric parts of the grid, but after that it was surprisingly smooth.  Not the ugliest looking puzzle in the world either.

Solved in 1'59.  A smooth enough start to the puzzle, it got a bit fiddly at the end.  Some interesting topological deductions in play here but it did feel somehow artificial.

Solved in 4'29.  I didn't want to have to guess to solve any of these puzzles, but I ended up doing so here.  There's probably a clever topological argument at play here, but this one definitely feels the most computer generated puzzle so far.

Solved in 3'13.  Again needed to guess to finish things off at the end, so pretty hard.  You can kind of see how I tried to build up a theme, and then compromised it quickly in order to get out the unique solution.  Not amazing either.

Solved in 1'54.  This one I remember trying to set by including a few patterns of clues I'm familiar with.  It kind of works, but not entirely successfully.  Definitely better than the previous two, but still with an undeniable artificial feeling to it.

Solved in 1'36.  This one is largely easy and with a smooth solve, except for a fiddly end which kind of ruins things a bit.  Again you can see how the initial visual theme got watered down a bit.  I suspect it wouldn't take that much work on this to keep the theme stronger and have a more consistent solve.

Solved in 1'20.  Basically a taburega puzzle for those familiar with the reference.  Solves smoothly and with a couple of interesting moments, but always with the sense that the thing is one big gimmick.

Solved in 2'16.  This one felt very computer generated.

Solved in 0'29.  Nice to have an easy one, although that was always going to be the case using patterns like this.  I suppose if I'm honest I'd have to say this one is fairly taburega-like as well.

Contact Form


Email *

Message *