Thursday, 28th March 2024
Puzzles Solved Yesterday: 124
Forum Index
 
Page 2 of 5<12345>
Masyu please!
Para
Kwon-Tom Obsessive
Puzzles: 1923
Best Total: 19m 28s
Posted - 2010.06.06 11:22:45
I understand what you mean Brian about the loops. But I guess it's just a matter of being used to a certain implementation. I've been solving all kinds of loop puzzles in square grids for years that I don't have any trouble seeing this compared to a slitherlink loop. Guess it's just a matter of experience. I just prefer the square grids because it's easier to judge all rules for me.

Nice puzzle btw.
Tilps
Kwon-Tom Obsessive
Puzzles: 6482
Best Total: 20m 22s
Posted - 2010.06.06 11:56:15
My silverlight version is making a bit more progress, puzzle generation is still to slow to recommend anything above 7x7, but the puzzles are a tad bit more interesting without having to invoke the 'unlimited' difficulty level.
mathmaniac
Kwon-Tom Obsessive
Puzzles: 1293
Best Total: 20m 57s
Posted - 2010.06.06 18:00:31
I'd say its working pretty well so far. Although they are not very difficult. Even the unlimited ones. Anything you can do to up the ante?
Tilps
Kwon-Tom Obsessive
Puzzles: 6482
Best Total: 20m 22s
Posted - 2010.06.07 01:10:42
I suspect that increasing puzzle size is key to increasing the probability of the puzzle being hard.  I've got a new idea for improving puzzle generation speed which I will implement tonight - hopefully it should bring generation performance close to LoopDeLoop and with that some serious puzzles.

If you were not specifying the size, the default size is 5, which definitely won't get you any hard puzzles.  Try it with 7, or if you are patient, 8.
MondSemmel
Kwon-Tom Obsessive
Puzzles: 6142
Best Total: 7m 47s
Posted - 2010.06.07 01:20:59
Quote:
Originally Posted by Tilps
I suspect that increasing puzzle size is key to increasing the probability of the puzzle being hard.  I've got a new idea for improving puzzle generation speed which I will implement tonight - hopefully it should bring generation performance close to LoopDeLoop and with that some serious puzzles.

Tilps, I was wondering - on a slightly different topic - whether it would be possible to decrease generation speed for beast size (40x30) slither link puzzles with LoopDeLoop. Currently it's at several hours (with all options turned on except for "Consider cell intersection interactions", "Generate puzzles with no multiple loop solutions" and "Use Cell Pairs inside trials"), and once a puzzle is finished, rating the puzzle (I have no idea what that means, by the way) takes even more time.
(Please take the above as hopefully constructive criticism - the reason I'm mentioning it is because I've used your LoopDeLoop program to generate more than 70 beast size puzzles in the past 9 months ...)

On a slightly different note, I should probably post these beast size puzzles in the forum, again.
Tilps
Kwon-Tom Obsessive
Puzzles: 6482
Best Total: 20m 22s
Posted - 2010.06.07 01:58:30
It took quite a bit of work to get the performance to where it is .  Unfortunately 'hard' beast puzzles take a long time to generate.  Easy ones are much faster, but I guess that doesn't really interest you
If you haven't already checked, make sure you are using the latest version, when I last worked on it the performance improved a bit in the final couple of versions. (I should probably also double check I actually released the latest version...)

By the way, if rating the puzzle doesn't interest you, just try and pretend its not there for now - its mainly so that (for someone who understands the shorthand) you can work out whether the puzzle generated is actually hard or not.  Since the generation only gaurantees that a puzzle will not be harder than a certain level, sometimes you can end up with a puzzle which is actually much easier than your target, and you'll want to regenerate.  Usually only applies for smaller puzzles though, wouldn't want to regenerate a beast...
Last edited by Tilps - 2010.06.07 03:13:23
Brian
Kwon-Tom Obsessive
Puzzles: 4739
Best Total: 9m 6s
Posted - 2010.06.07 09:32:49
Quote:
Originally Posted by Para
I understand what you mean Brian about the loops. But I guess it's just a matter of being used to a certain implementation. I've been solving all kinds of loop puzzles in square grids for years that I don't have any trouble seeing this compared to a slitherlink loop. Guess it's just a matter of experience. I just prefer the square grids because it's easier to judge all rules for me.
I don't think I've done Masyu before, so that might be part of it. And I can understand the need for square grids with some other loop puzzles--but imagine doing slitherlink with a grid in the background! Anyway, I've said enough already, so I'll drop the subject.

Quote:
Nice puzzle btw.
Thanks. Here's another one:

Tilps
Kwon-Tom Obsessive
Puzzles: 6482
Best Total: 20m 22s
Posted - 2010.06.07 11:16:46
I had a false start implementing my faster puzzle generation algorithm, it barely got me one size larger for the same speed, but I found the bugs and it now generates puzzles much faster.

(In case anyone cares, the algorithm is to identify edges in a generated loop, which are not affected by any clue, then see if you can move the loop segment which contains them without moving any edge which is affected by a clue.  One or two iterations of that seems to increase the probability of the loop being solvable a huge amount.)

I've raised the default size to 7 for now, but even 15x15 puzzles are usually not a long wait.

I also increased the difficulty spread before jumping to unlimited. Moderate and Maybe Harder should be a tiny bit more challenging.
Jankonyex
Kwon-Tom Obsessive
Puzzles: 5663
Best Total: 9m 35s
Posted - 2010.06.07 14:17:37
[jmy01]                                                                                   difficulty: easy

Last edited by Jankonyex - 2010.06.09 01:41:31
Brian
Kwon-Tom Obsessive
Puzzles: 4739
Best Total: 9m 6s
Posted - 2010.06.07 19:21:35
Quote:
Originally Posted by Jankonyex
[jmy01]
Ah, color deductions. Very nice. I had these in mind when I started constructing my first puzzle, but it ended up not needing them.
Brian
Kwon-Tom Obsessive
Puzzles: 4739
Best Total: 9m 6s
Posted - 2010.06.08 09:52:35
Another. This one's probably on the hard side, at least for 9x9. I don't have much experience creating puzzles, but it's kind of fun. I think I only made one or two slitherlinks by hand, the rest were with a program.

Jankonyex
Kwon-Tom Obsessive
Puzzles: 5663
Best Total: 9m 35s
Posted - 2010.06.08 20:16:38
2 similar puzzles (rated as easy because no FP is needed):

[jmy02a]                                                                                              difficulty: easy


[jmy02b]                                                                                              difficulty: easy


hint:


both are impossible

don't like Masyu very much ... because of its strange rules
Last edited by Jankonyex - 2010.06.08 20:16:55
Brian
Kwon-Tom Obsessive
Puzzles: 4739
Best Total: 9m 6s
Posted - 2010.06.09 02:38:40
A few more.





Jankonyex
Kwon-Tom Obsessive
Puzzles: 5663
Best Total: 9m 35s
Posted - 2010.06.09 05:05:20
pattern:

Brian
Kwon-Tom Obsessive
Puzzles: 4739
Best Total: 9m 6s
Posted - 2010.06.10 05:15:43
foilman,

I was wondering how difficult it would be to adapt your flash solver to Masyu. I'm not talking about puzzle generation, only about solving and verifying a solution. The layout would be almost exactly the same, but with no numbers and some of the dots as white or black circles. The process of line drawing, crosses, shading and fixed position would be the same. And as far as verifying solutions, you'd only need to check that the rules are followed at the white/black circles and that there's only one loop. (Puzzles could still have multiple solutions, but that would be the fault of the puzzle creator, not the flash applet.)

Then, instead of trading images here we could trade links, like we do for slitherlink puzzles:
  http://www.kwontomloop.com/puzzlef.php?loopy=10x10:33b2....

Would that take a lot of effort on your part? I think your implementation of slitherlink is by far the best out there, and I think it would work perfectly for Masyu.

Also, congratulations on the baby, if that's happened yet.
Tilps
Kwon-Tom Obsessive
Puzzles: 6482
Best Total: 20m 22s
Posted - 2010.06.10 07:02:34
Twist for Silverlight 3
I haven't tried this, but I think this one might be moonlight compatible (although you may need to get the moonlight 3 preview from http://www.go-mono.com/moonlight/prerelease.aspx). 
Compared to the main twist version, this one doesn't support right click, otherwise they should be practically the same.
Tilps
Kwon-Tom Obsessive
Puzzles: 6482
Best Total: 20m 22s
Posted - 2010.06.10 08:24:59
A Puzzle (Silverlight 4)
A Puzzle (Silverlight 3)

I invented a 'loopy' style notation for puzzles. 0 means the empty circle, 1 means the filled circle, letters correspond to the number of 'normal' intersections.
(Note that I went with width and height in terms of dots, where my game actually currently counts squares (legacy left over from being based on LoopDeLoop).
Another puzzle (SL4)
A larger puzzle (SL4)
Last edited by Tilps - 2010.06.10 08:52:34
Brian
Kwon-Tom Obsessive
Puzzles: 4739
Best Total: 9m 6s
Posted - 2010.06.10 10:22:26
Ah, that works! Thanks, Tilps. It's pretty nice.

One bug/request, though. Whenever I finish a puzzle it automatically puts up a new puzzle (it also pauses for a few seconds before drawing the final line). I'd rather have to click "generate" to get a new puzzle.
Tilps
Kwon-Tom Obsessive
Puzzles: 6482
Best Total: 20m 22s
Posted - 2010.06.10 13:29:13
The auto generate next pre-dates the generate button existing. I actually like it, but I can certainly make it optional.  I need to do several things to make it more user friendly, it is little more than a proof of concept experiment in programming for silverlight at the moment.
Not to mention that the autogenerate on complete is rather broken if you follow a puzzle link like in my previous post.

Btw, the delay adding the last line is proving the puzzle is correctly solved.  It does that by actually solving the puzzle and comparing results, not by verifying the current loop satisfies the constraints. (So if the puzzle actually has multiple solutions, it will never mark you as having finished.)

Edit: For the time being I have simply removed the autogenerate.
Last edited by Tilps - 2010.06.10 14:30:46
Jankonyex
Kwon-Tom Obsessive
Puzzles: 5663
Best Total: 9m 35s
Posted - 2010.06.10 14:01:52
jmy03
jmy04

solution
Page 2 of 5<12345>

Forum Index