| New puzzle generator |
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.11.15 08:58:16
Quote: Originally Posted by mrtheman Hmm, some interesting reading here. A question though; do you think it is actually worth implementing all those advanced deducing rules into a program? I have also written a slither link generator/solver which does do guessing, and although it may be slower, it certainly generates puzzles much harder than the ones you posted here.
Also, with adding all these patterns, aren't you also making it think more like a human, which results in the puzzles generated being actually easier for humans to solve?
I do get that it's just good fun to write a generater that is very advanced, though I'm not sure this is the way you will be creating the best puzzles..
That's just my view though, other opinions are much appreciated. |
As I said at then beggining - the goal of this project was to write generator with very good ratio difficulty/speed. It is obvious that extensive use of trying edges can leads to more difficult puzzle, but there is difference between 2 mins for generating beast-sized puzzle (with difficulty nearly same as KT beasts) and hour and more with only backtrack..
There are some techniques that aren't similar to human deductions (about 5 techniques).
I think that for usual player puzzles are hard enough. I didn't want to generate the best puzzles (in terms of difficulty).
And the last thing - I tried to found out any puzzle of you (as you mentioned), but I have not found at least one. So I could not compare your puzzles.. |
Para
Kwon-Tom Obsessive
Puzzles: 1923
Best Total: 19m 28s | Posted - 2009.11.15 20:22:56
I like the idea of a technique based solver/generator much more than backtracking solver/generators. It leads to nicer puzzles. Just because a generator creates more difficult puzzls doesn't mean it's better. It's the same with sudoku generators. The best generators are the technique based programs. Especially if you can select which techniques you want in your puzzle and which you don't want. This way you know the techniques used are ones you find acceptable/like to use. The longer you have been posting puzzles in the thread, the more in enjoy the logic in them. |
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.11.15 20:46:24
I share the same idea as Para.. I have now implemented 7 levels (and I think that it is all I was able to do) - from very, very beginers (only satisfying clues and dots constrains - maybe the same difficulty level as Loopy easy), then easy level (only basic locks), and more and more advanced techniques later.. So the user will be able to choose the adequate level which fits him the best.. In fact there are 14 levels, because in each level category you can select if you also want to use shading techniques (which also differs from level to level).. So it is for beginers, intermediate users and also for experienced users - each can choose the appropriate level.. And it is really fast.. |
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.11.17 11:38:53
One bigger (20x14) with all features switched on - posted just for fun.. Puzzle |
MondSemmel
Kwon-Tom Obsessive
Puzzles: 6159
Best Total: 7m 47s | Posted - 2009.11.17 17:07:03
Quote: Originally Posted by v_e_e_n_c_a One bigger (20x14) with all features switched on - posted just for fun.. Puzzle |
Solved in 8m6s. Good puzzle - I liked it.
EDIT: I'm not sure how much easier it would have been by using highlander deductions, though.
Last edited by MondSemmel - 2009.11.17 17:07:24 |
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.11.19 08:16:33
Can anyone tell me some tips how to solve this puzzle (generated by my solver) Puzzle to solve ?
The lack of clues in the middle probably leads to highlander deduction (with which I am not still familiar). So I will be glad for any suggestions. Here is the situation which I achieved (without trying poss.)
Last edited by v_e_e_n_c_a - 2009.11.19 08:19:11 |
chairman
Kwon-Tom Obsessive
Puzzles: 1397
Best Total: 17m 32s | Posted - 2009.11.19 08:44:09
To avoid a local loop, there must be a line at the right hand side of the third row from top. |
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.11.19 08:47:53
Quote: Originally Posted by chairman To avoid a local loop, there must be a line at the right hand side of the third row from top. |
Thanks, I forgot it althought I have implemented this technique.. 
Edit: But still not able to solve the middle without clues..
Last edited by v_e_e_n_c_a - 2009.11.19 08:49:22 |
Tilps
Kwon-Tom Obsessive
Puzzles: 6722
Best Total: 18m 37s | Posted - 2009.11.19 12:06:46
The not closing the loop keeps going for a bit, and that with the diagonal feed down to the three pair almost gives the bottom right corner area. There is a horizonal edge above the second bottom row which if closed causes an area with 3 input lines. I suspect some kind of advanced shading/counting will prove it a cross, but it was just obvious. From there it all pretty much fell out.
Last edited by Tilps - 2009.11.19 12:07:28 |
chairman
Kwon-Tom Obsessive
Puzzles: 1397
Best Total: 17m 32s | Posted - 2009.11.19 12:18:57
I was just about to write a similar thing, but Tilps was ahead of me.
Let me reply to another post of you. The only puzzle of MrtheMan I know of on this site is here. It is a long time ago when I solved it, I will try it again soon to compare it with yours.
It would be nice to have the possibility to keep the creator of an uploaded puzzle secret (for a week or so), so the solvers could guess who made it. |
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.11.19 12:48:48
Quote: Originally Posted by chairman I was just about to write a similar thing, but Tilps was ahead of me. Let me reply to another post of you. The only puzzle of MrtheMan I know of on this site is here. It is a long time ago when I solved it, I will try it again soon to compare it with yours. It would be nice to have the possibility to keep the creator of an uploaded puzzle secret (for a week or so), so the solvers could guess who made it. |
I saw that you had been beta-tester of his app.. What were times for generating larger puzzles? I read that it was possible up to 50x50.. Weeks before I tested my generator and it is not problem to generate 128x128 and bigger (but needs more time).. I didn't posted this size because there wasn't any app to see whole puzzle..
That feature would be nice, but you have to contact Foilman probably..
Last edited by v_e_e_n_c_a - 2009.11.19 12:52:23 |
Tilps
Kwon-Tom Obsessive
Puzzles: 6722
Best Total: 18m 37s | Posted - 2009.11.20 00:02:46
I just created a new puzzle by hand which I put in the archives - #414. I think it is kind of annoying for a solver, but easy for a human.
I was wondering if anyone's solver can solve it without doing a full recursive attack?
Last edited by Tilps - 2009.11.20 00:03:36 |
Brian
Kwon-Tom Obsessive
Puzzles: 4907
Best Total: 9m 6s | Posted - 2009.11.20 02:28:52
Quote: Originally Posted by v_e_e_n_c_a Can anyone tell me some tips how to solve this puzzle |
|
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.11.20 09:07:29
Thanks to all who posted me tips or solutions..
Interesting puzzle Tilps, I realized that althought I have implemented some advanced path techniques it is still not able to solve it.. Good situation to think about..
And one question - what do you think about design of board (image i have posted)? I mean if the line segments are easily visible althought the shading is used and so on.. I'll be glad for any opinion..
Last edited by v_e_e_n_c_a - 2009.11.20 09:07:41 |
Tilps
Kwon-Tom Obsessive
Puzzles: 6722
Best Total: 18m 37s | Posted - 2009.11.20 09:44:52
Quote: Originally Posted by v_e_e_n_c_a As I said before my solver don't use backtrack and from this reason it is very fast.. I tried to compare also difficulty with other slither link apps..
E.g.: for this puzzle (randomly generated with my program with advanced logic switched on)
10x10:c11b2b2a3a1212a22a1b1233d2a2d3b0a12a21a2a1e2b11c2b31b2c33b1a23a1c2a32c2a
Loopy : only four crosses were found (around 0)
LoopDeLoop : four crosses found (around 0) and one cross between 1 & 1 up.. Nothing else without trying possibilities.. (Rated as SI10 - Depth:6 and S - UnFin:82%) |
I felt inspired by this to add another small feature to LoopDeLoop.
In the current unreleased edition (unreleased because I need to add yet another flag to make this new feature optional) with all options enabled LoopDeLoop solves half of this puzzle using autostart alone.
Have to say I can't seem to solve the rest myself without a single trial. (My solver doesn't have advanced path deductions, but I can't see them being useful until the very end...)
Last edited by Tilps - 2009.11.20 10:04:44 |
Tilps
Kwon-Tom Obsessive
Puzzles: 6722
Best Total: 18m 37s | Posted - 2009.11.20 10:53:01
I was reading through some of the forum archives trying to find something when I saw a veenca post mentioning local contradiction in respect to solvers.
I was wondering if your solver still has a local contradiction component - since that is a significant part of what I call a trial. (Admittedly my definition of 'local' is flexible...) |
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.11.20 12:06:14
Quote: Originally Posted by tilps I was reading through some of the forum archives trying to find something when I saw a veenca post mentioning local contradiction in respect to solvers.
I was wondering if your solver still has a local contradiction component - since that is a significant part of what I call a trial. (Admittedly my definition of 'local' is flexible...) |
I don't know which post do you mean (lazy to find it), but I don't use local contradiction component. Or to be more precious - not in terms choosing some edge, setting some value (cross or line) and making deductions to find contradictions. It deals with locks and permitted states - eg. suppose 3 3 pattern (diagonal), then 2 edges can't be both lines in the dot they share, because it will break the clue constraint of 3 (only one cross allowed). So it is not trial - it is deduced automatically without trying edges. It deals more with cells (clues) interaction and it is made in constant time.. |
chairman
Kwon-Tom Obsessive
Puzzles: 1397
Best Total: 17m 32s | Posted - 2009.11.21 10:07:56
Quote: Originally Posted by v_e_e_n_c_a
I saw that you had been beta-tester of his app.. What were times for generating larger puzzles? I read that it was possible up to 50x50..
|
I can't tell you that. There was a fixed test set, in my memory consisting of +-20x14 sized puzzles, ranging from easy as Monday puzzles to very hard, probably comparable to Naivoj's #411 (still did not solve that one). I was not able to solve all puzzles before the testing period was over. |
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.11.24 09:22:23
I improved better propagation and interaction between rules. It can make puzzles harder (with no much more time needed). I use it only in the highest level of difficulty - to stay other leves easier. I am posting here some (small) puzzles - you can try them and write your opinion about puzzles. I also added one beast (in other thread of forum), so you can also try that..
Here are some puzzles:
Puzzle 1
Puzzle 2
Puzzle 3
Puzzle 4
Puzzle 5 - symmetrical - easier |
v_e_e_n_c_a
Kwon-Tom Obsessive
Puzzles: 2080
Best Total: 32m 53s | Posted - 2009.11.24 12:17:59
I would be glad if you post me how difficult these puzzles were for you.. And if they were solvable without fixed positions for you.. I'll be also pleased if Foilman could post me ratings (made with his system) for these puzzles.. Thanks for any comments.. |
