| New puzzle generator |
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.04.21 16:47:28
I have to apologize.. Last set which I provided was not generated with shading system what I promised (in generator it was switched off) - it was my mistake.. So here is a set of puzzles with this feature turned on..
http://kwontomloop.com/puzzlef.php?loopy=10x10:c2b2322a2c31c212a222b1a2d221122a3a1a2a2c22b0b3a32b21a2b2b2e31d2c23b02c
http://kwontomloop.com/puzzlef.php?loopy=10x10:a2a13b3b3a3b2c23a222313d221a1i222d2b1b2a2b221a1a3c2a1b22a2a111b3a3a2b231a
http://kwontomloop.com/puzzlef.php?loopy=10x10:b231a3b012b2a1121a3a3a2f1a2213a32h3a212a3a2c2b3b0a21d2e2b3a2a233b2a2b2a |
Jankonyex
Kwon-Tom Obsessive
Puzzles: 5680
Best Total: 9m 35s | Posted - 2009.04.21 17:53:57
Quote: Originally Posted by v_e_e_n_c_a ...here is a set of puzzles with this feature turned on.. |
(some hidden numbers are shown for the sake of simplicity)
 |
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.04.21 21:00:47
Quote: Originally Posted by jankonyex Quote: Originally Posted by v_e_e_n_c_a ...here is a set of puzzles with this feature turned on.. |
With the help of highlander, I solved the first two within 2 minutes and the last one within 1 minute. Without it, how to solve the first one at this stage by coloring rules? (some hidden numbers are shown for the sake of simplicity) |
Ok, at first one of very basic rule is that 2 cells connected with cross has to have same color and 2 cells connected by line has to have opposite color. (Rule 1)
Next important thing is that every clue 2 has on his "cross" (4 cells around this clue) two same and two opposite colors (as a proof just draw all combinations of lines and crosses and than look for the colors on his "cross" according to Rule 1) - call this eg. Rule 2
So due to Rule 1 - you can see two colors - green and blue on the picture. And according to Rule 2 - you have two opposite colors (green and blue) and one known color (red) around clue 2 - to satisfy Rule 2 cell A has to have opposite color to red => so it has to be colored yellow.
Now if you look to clue 1 just under cell A, you can see that on "cross" it has two opposite colors now (red and yellow). And two opposite colors around 1 implies antilock (just that one edge has to be cross and the other line) -> so it implies that you can simply deduce edges marked as d on the picture. And I believe that from this point you can make a lot of deductions..
Hope that it will be usefull..
Last edited by v_e_e_n_c_a - 2009.04.21 21:03:27 |
Jankonyex
Kwon-Tom Obsessive
Puzzles: 5680
Best Total: 9m 35s | Posted - 2009.04.22 01:51:36
Quote: Originally Posted by v_e_e_n_c_a so it implies that you can simply deduce edges marked as d on the picture. |
Last edited by Jankonyex - 2009.04.22 02:11:45 |
Jankonyex
Kwon-Tom Obsessive
Puzzles: 5680
Best Total: 9m 35s | Posted - 2009.04.22 02:33:20
what about this:
 |
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.04.22 09:58:43
I want to say something about difficulty of puzzles.. I think that difficulty of puzzles is very subjective thing based on rules (patterns) known to the (human) solver.. Many of generators use their fixed rule set or very simple deduction rules.. So when you get familiar with rules from the current generator, than puzzles are easy for you and I think that they can become boring.. Just you see 33 pattern and you can draw lines and crosses.. If someone want to generate harder puzzle, he must use backtrack which will result in hiding some other clues.. But my solver (and generator) doesn't use any fixed rule set.. It just deduce rules in time of generation and solving (without backtrack).. Many people (as me) know only very small patterns (as 33 rule and so on), but patterns should be bigger and their count is infinity - if you know all patterns you will be able to solve all puzzles immediatelly (just because every the puzzle is big patter in fact). It is also the reason why my 1-2 clues beast was too difficult - it is only because that there are no many well known patterns of 1 and 2. I think that it is advantage that my generator doesn't use fixed sized rule set - just because it can generate new and new patterns and it shouldn't become boring.
For example suppose one situation - it could be also called pattern - you can deduce 4 edges.
|
MondSemmel
Kwon-Tom Obsessive
Puzzles: 6159
Best Total: 7m 47s | Posted - 2009.04.22 10:17:52
Your puzzles aren't bad, but if you want to make them more difficult, you really have to decrease the amount of highlander patterns (they make (some of) your puzzles far easier than they would otherwise be). |
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.04.22 10:33:48
Quote: Originally Posted by mondsemmel Your puzzles aren't bad, but if you want to make them more difficult, you really have to decrease the amount of highlander patterns (they make (some of) your puzzles far easier than they would otherwise be). |
I don't use any of highlander patterns in generation.. It is just a method which help you to solve harder puzzles.. I am sure that puzzles would be more difficult when I finish my work - it is still in progress.. I have implemented only clue constrains and coloring yet (which made puzzles a little bit harder I think) Now I am working on path deductions and then regions deduction, so I hope that it would increase difficulty of puzzles.. But times for generation will be slower - eg. beasts without coloring were generated in 14-16 sec, now with use of this method times increased to 20-22 sec.. |
MondSemmel
Kwon-Tom Obsessive
Puzzles: 6159
Best Total: 7m 47s | Posted - 2009.04.22 10:50:56
Quote: Originally Posted by v_e_e_n_c_a I don't use any of highlander patterns in generation.. It is just a method which help you to solve harder puzzles.. I am sure that puzzles would be more difficult when I finish my work - it is still in progress.. I have implemented only clue constrains and coloring yet (which made puzzles a little bit harder I think) Now I am working on path deductions and then regions deduction, so I hope that it would increase difficulty of puzzles.. But times for generation will be slower - eg. beasts without coloring were generated in 14-16 sec, now with use of this method times increased to 20-22 sec.. |
I'll try to clarify what I meant. Nobody ever uses highlander patterns when generating puzzles (if you did, your puzzles might no longer have a unique solution), it's a technique used to solve puzzles based on the uniqueness of a puzzle combined with a LACK of clues. Usually you try to minimize the amount of clues when generating to make puzzles harder, but if this leads to highlander patterns, it makes puzzles easier. In other words, highlander patterns are the only case where adding clues might make a puzzle more difficult instead of easier. Whether you get rid of the highlander patterns by adding clues or by preventing these patterns doesn't matter, but if you do neither, your puzzles will regularly end up with highlander patterns (2 in a corner etc.).
Last edited by MondSemmel - 2009.04.22 10:52:10 |
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.04.22 10:58:38
to MondSemmel: I think that I understand you now.. Is there any topic on this forum with list of highlander rules or some detailed discussion on this topic? I want to learn more about that to produce better puzzles..  |
MondSemmel
Kwon-Tom Obsessive
Puzzles: 6159
Best Total: 7m 47s | Posted - 2009.04.22 11:17:00
Quote: Originally Posted by v_e_e_n_c_a to MondSemmel: I think that I understand you now.. Is there any topic on this forum with list of highlander rules or some detailed discussion on this topic? I want to learn more about that to produce better puzzles.. |
Heh. Déjàvu  . Guess who created a thread about this very topic?^^
http://kwontomloop.com/forum.php?a=topic&topic_id=341
A few are also in this thread:
http://kwontomloop.com/forum.php?a=topic&topic_id=283 |
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.04.22 11:20:29
Quote: Originally Posted by mondsemmel Heh. Déjàvu . Guess who created a thread about this very topic?^^ http://kwontomloop.com/forum.php?a=topic&topic_id=341 |
 I know but I didn't get familiar with this technique yet.. And there was not enought informations for me.. But thanks for second link.. |
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.04.22 18:40:32
And what about this puzzle? Is there a lot of highlanders or is it so easy? I think that one is in left top corner (but not sure about them)..
http://kwontomloop.com/puzzlef.php?loopy=10x10:a2222b11a2e2h3a3233a2a221b2a212b1a3c32a21a2a3a2b1c322a23a2a32c2222b23b2232b3
Edit: I am sorry, I just can't see one well known pattern - so it is easy..
Last edited by v_e_e_n_c_a - 2009.04.22 18:43:48 |
Tilps
Kwon-Tom Obsessive
Puzzles: 6720
Best Total: 18m 37s | Posted - 2009.04.22 23:17:42
Quote: Originally Posted by v_e_e_n_c_a
I have implemented a little bit more.. Imagine that you have clue 1, and cell up to thic clue is red and cell under this clue is yellow. Then you can deduce 2 crosses, which is not possible from locks/antilock..
|
The two edges on the 1 are antilocked - antilocked interacts with the 1 to prove two crosses, since one of the antilocked edges is filled in and only 1 edge total can be filled in.
Antilocked/locked edges are frequently not touching, being across a cell is quite common, sometimes they are even remote.
Specifc example where they are remote is shading has proven two remote cells have related colors, and also a cell adjacent to each of those two cells have related colors. But the two sets are not related to each other. In that case the edges between the adjacent pairs of cells are locked or antilocked.
Last edited by Tilps - 2009.04.22 23:28:49 |
Tilps
Kwon-Tom Obsessive
Puzzles: 6720
Best Total: 18m 37s | Posted - 2009.04.22 23:35:13
Path deduction was mentioned.
Does anyone have a good algorithm for proving the question marks are crosses here without invoking multiple simultaneous trials?
I've been thinking about this for a while but have had no luck coming up with something which doesn't potentially devolve into an exponential number of states being considered during a trial. |
Brian
Kwon-Tom Obsessive
Puzzles: 4907
Best Total: 9m 6s | Posted - 2009.04.23 02:56:02
These are nice puzzles, v_e_e_n_c_a, although I haven't done all of them. I've actually been rewriting my generator (which I hadn't touched in two years). It's much simpler now and only takes a few seconds to generate larger puzzles. I'm not happy with the difficulty yet, but I'll probably post some puzzles later this week.
Edit: Well, here's one.
Last edited by Brian - 2009.04.23 03:00:36 |
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.04.23 05:55:18
Quote: Originally Posted by tilps Antilocked/locked edges are frequently not touching, being across a cell is quite common, sometimes they are even remote. |
I am sorry, but it is very common that if somebody implements locks, he uses only interaction of edges joined by common dot - no remoted interactions and so on - in that case it wouldn't be able to deduce two crosses in example I have posted.
I use also some remoted deductions in coloring and in basic constrains. |
Brian
Kwon-Tom Obsessive
Puzzles: 4907
Best Total: 9m 6s | Posted - 2009.04.30 00:29:14
#398 solution, very nice. |
Jankonyex
Kwon-Tom Obsessive
Puzzles: 5680
Best Total: 9m 35s | Posted - 2009.04.30 10:30:10
Quote: Originally Posted by brian |
solution
including a very trivial highlander lemma that everyone can figure out. |
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.05.01 08:10:48
Does anybody know if there exists a way how to convert puzzles from archive to Loopy string without rewriting it? It is only for testing weak sides of my solver (what I should improve) - for path deduction and region analysis (I just want to know what my solver is able to solve and if there is something to deduct (without backtrack). I found weakness in path deduction and preventing small loops, which are obvious for human - and I finally made up an alogorithm which I am implementing now (based on graph theory). I saw that Brian and Jankonyex made solution for puzzle I added to archive in Loopy so I am interested if there is a way how to capture puzzle without rewriting it. |
