Discussions on User Beast Puzzles |
MondSemmel Kwon-Tom Obsessive Puzzles: 6159 Best Total: 7m 47s | Posted - 2008.11.07 23:01:46 Anyone...? |
lodenkamper Kwon-Tom Fan Puzzles: 21 Best Total: 47m 58s | Posted - 2008.12.03 07:33:05 I just posted some new user beasts, but I don't know how to add flash links. If someone can tell me how (or point me to an explanation), I'll go ahead and add flash links for these user beasts |
MondSemmel Kwon-Tom Obsessive Puzzles: 6159 Best Total: 7m 47s | Posted - 2008.12.04 23:29:08 Thanks a lot for the puzzles, lodenkamper! I've been trying to solve them without highlander patterns (which help dramatically in all big puzzles) and haven't even solved a single one yet. I guess if I become too frustrated at the more obvious highlander situations, I'll just use highlander again, but until then, the puzzles definitely are very challenging, but their difficulty may or may not tremendously decrease if one applies highlander patterns. On the note of Flash links: I thought it would be more difficult, but it's actually really just adding the different loopy strings at the end of http://www.kwontomloop.com/puzzlef.php?loopy=
For example, the URL for Naivoj's #52 is http://www.kwontomloop.com/puzzlef.php?loopy=40x30:3a3b2d213d1a23a3b3a2c1a2a221a31c2a221b3a3... (abbreviated so it doesn't break the forum format)
which is simply said code plus the loopy string. You should probably use the forum tags ['url=LINK]Flash Beast #X['/url] (without the ''s) to shorten them, or the Flash links will be too long^^. Thanks again! |
MondSemmel Kwon-Tom Obsessive Puzzles: 6159 Best Total: 7m 47s | Posted - 2008.12.08 18:11:55 I've solved all the new puzzles now - the hardest one was probably 102. I didn't manage to solve a single one without highlander patterns, though^^...but they all became substantially easier when I used the highlander principle. Without that, though, some of them seemed pretty unsolvable. Great job! You may not have a good way to rank their difficulty, but from my perspective, I'd call them quite to very (102) challenging. Thanks again, I really enjoy solving these beasts! And if you ever feel like posting more - I'll never say no to more of them, ever :p. |
MondSemmel Kwon-Tom Obsessive Puzzles: 6159 Best Total: 7m 47s | Posted - 2009.03.10 21:56:53 I just wanted to say "thanks" for the new user beasts - I've solved them already, but they were really tough (and I really like tough puzzles ). Am I right in my estimation that #106 was the hardest of these puzzles? It was also the last or second to last puzzle I solved, so that might have biased my perception, but solving that one took tons of extremely small steps before I could finally make some real progress.
If you ever feel like posting more, I'll be sure to solve them =) - these beasts are almost like an addiction^^. Thanks again! On that note, if anybody has a specific question regarding any of the beasts on this site (beasts of the months or user beasts), I've probably done all of them often enough to be in a good position to help . |
lodenkamper Kwon-Tom Fan Puzzles: 21 Best Total: 47m 58s | Posted - 2009.03.11 06:13:33 You're welcome. Thanks for the feedback on difficulty. The time to solve for my solver is roughly equal for 105-109, but solving time is not a reliable measure of difficulty as seen by a human solver. |
MondSemmel Kwon-Tom Obsessive Puzzles: 6159 Best Total: 7m 47s | Posted - 2009.03.13 19:02:44 Thanks again for the puzzles . I'm sorry I can't say with a lot of certainty how difficult this batch of puzzles was; I made mistakes in two of the puzzles, so I had to restart both of them (mistakes are so incredibly annoying...). Other than that, I had lots of unsuccessful trial & error explorations in 114, so I needed a long time before making real progress (and it is really satisfying to see a huge chunk of the puzzle solved when it took so long to get there). Other than that, I'd agree that the puzzles in this batch felt quite similar in difficulty. If you feel like posting more, I'll definitely feel like solving them . Oh, and as you mentioned that you have your own solver, if there's anything more specific you want to hear about solving your puzzles, tell me beforehand if you want and I'll keep notes or something^^. |
Tilps Kwon-Tom Obsessive Puzzles: 6722 Best Total: 18m 37s | Posted - 2009.04.11 14:55:14 Latest LoopDeLoop solving times in minutes and seconds: (3.0 Beta 38 )
Beast 18 - 1:20 Beast 73 - 5:00 Beast 88 - 1:10 Beast 88 Reverse - 0:27 Beast 99 - Multiple solutions detected?? (Edit: this was a bug, fixed in Beta 39)
Last edited by Tilps - 2009.04.18 15:58:34 |
lodenkamper Kwon-Tom Fan Puzzles: 21 Best Total: 47m 58s | Posted - 2009.04.16 02:52:09 It's good to see various people contributing new user beasts. Thanks. I should be able to put some more up shortly.
One recommendation --- it probably makes sense to keep going with the numbering started by Naivoj, so we can more easily refer to specific puzzles. |
lodenkamper Kwon-Tom Fan Puzzles: 21 Best Total: 47m 58s | Posted - 2009.04.16 03:05:58
Quote: Originally Posted by tilps Latest LoopDeLoop solving times in minutes and seconds: (3.0 Beta 38 )
Beast 18 - 1:20 Beast 73 - 5:00 Beast 88 - 1:10 Beast 88 Reverse - 0:27 Beast 99 - Multiple solutions detected?? |
Those are some nice solving times, especially #88 reverse.
According to my solver, user beasts 1-114 all have unique solutions. For beasts 1-99 posted by Naivoj, this amounts to independent confirmation of his results.
Does LoopDeLoop give a multiple solutions message when multiple solutions are confirmed (i.e., solver finds 2 distinct solutions), or when multiple solutions are suspected (i.e., no further inferences possible)? |
pqg Kwon-Tom Obsessive Puzzles: 6385 Best Total: 15m 37s | Posted - 2009.04.16 14:47:16 Thanks to all who've contributed to the recent flurry of new puzzles. I've solved them all and it's interesting to note some differences in style between them. Lodenkamper's were all a pretty consistent (relatively tricky) level.
Tilips rated the difference in difficulty of his set pretty accurately (I'd rate the first one as comparable to a slightly above-average difficulty Naivoj beast, while the last one is more like a slightly easier-than-average BotM) Might I suggest that you need a better measure of how much the loops fills the area than proportion of vertices visited? As you've already noted, that often leads to large areas of 0s but what you may not have picked up on is that it also leads to the non-zeroed area being filled very densely by often having big areas of 2's which end up being all parallel lines.
Although claimed to be all around the same difficulty, I found veenca's set to vary enormously. The first one (1&2 clues only) was definitely the toughest (followed by the 0,1,2 one). Despite your recommendation, I used FP and highlanders extensively. They highlighted a difficulty human solvers have with a certain type of puzzle that poses less difficulties for automated solvers - when a single deduction leads to a large number of simultaneous implications in different parts of the board, it's not easy to keep track of them all systematically, which can result in missing out a key x leaving you stuck later.
As regards the uniqueness of #99, I can only say that I solved it with a lot of highlander deductions, so the assumption that there is a unique solution certainly leads to one. (after the amount of time it took, I'll be interested if not a little frustrated if there does turn out to be multiple solutions...) By the reported times, veenca's solver appears to be among the fastest - have you tried running it on Naivoj's beasts? (in particular, 73,88,98 & 99)
Last edited by pqg - 2009.04.16 14:57:27 |
v_e_e_n_c_a Kwon-Tom Obsessive Puzzles: 2080 Best Total: 32m 53s | Posted - 2009.04.16 15:43:27
Quote: Originally Posted by pqg Thanks to all who've contributed to the recent flurry of new puzzles. |
If you enjoyed puzzles provided by me I can add some more puzzles.
Quote: Originally Posted by pqg Although claimed to be all around the same difficulty, I found veenca's set to vary enormously. The first one (1&2 clues only) was definitely the toughest (followed by the 0,1,2 one). Despite your recommendation, I used FP and highlanders extensively. They highlighted a difficulty human solvers have with a certain type of puzzle that poses less difficulties for automated solvers - when a single deduction leads to a large number of simultaneous implications in different parts of the board, it's not easy to keep track of them all systematically, which can result in missing out a key x leaving you stuck later. |
Yes, you are right.. But the difficulty differs in the clues set given - it is just caused by fact that 0 can gain the most information, from clue 3 we can gain also a lot of information and clues 1 and 2 gives much less information for solving - it just because there are much more possibilities how the loop can behave around these clues. So I have to introduce to my generator this fact - fraction of clues 0, 1, 2 and 3 which appears between clues. When I say that they are about same difficulty I meant the method used for creating them is just same. No backtrack only deductions.. And no advanced techniques which I want to add.. And I agree that for human solver it should be very hard to keep the track of changes in puzzle and implications.
Quote: Originally Posted by pqg As regards the uniqueness of #99, I can only say that I solved it with a lot of highlander deductions, so the assumption that there is a unique solution certainly leads to one. (after the amount of time it took, I'll be interested if not a little frustrated if there does turn out to be multiple solutions...) By the reported times, veenca's solver appears to be among the fastest - have you tried running it on Naivoj's beasts? (in particular, 73,88,98 & 99) |
My solver isn't able to solve all lines of other beasts - it's just because I have now implemented only basic deductions. No backtrack and so on (and I have to implement this - just because SlitherLink is well known to be NP complete - but I wouldn't use it in my generator - only in my solver for user provided puzzles) So when I will finish this project, it should be iteresting what times I will give for them.
Solving times for my beasts with my solver:
1-2 Clues: 0.0468 seconds 2-3 Clues: 0.0624 seconds 0-1-2 Clues: 0.0468 seconds 1-2-3 CLues: 0.0468 seconds and regular ones also 0.0468 seconds
Last edited by v_e_e_n_c_a - 2009.04.16 15:45:17 |
MondSemmel Kwon-Tom Obsessive Puzzles: 6159 Best Total: 7m 47s | Posted - 2009.04.16 21:11:23 Thanks again for the puzzles. I always like Naivoj's and lodenkamper's puzzles (due to their difficulty) but veneca's were quite interesting, too - limiting puzzles to certain times of numbers certainly makes them vary a lot in difficulty but also makes them quite interesting. If anyone of you requires any specific information about solving these puzzles, just tell me - except for Tilps' puzzles, I've solved them all, I think (I screwed up in one of his puzzles and was annoyed for a while - I'll come back to them later). I think veneca's 2-3 clues puzzle is probably the most educational of the bunch - there aren't that many patterns for 3s and 2s which makes it comparatively easy to solve, and the size of the puzzle allows for a lot of variety in these patterns. Please post more (especially Naivoj's and lodenkamper's difficult ones). I'll definitely solve them^^. |
Tilps Kwon-Tom Obsessive Puzzles: 6722 Best Total: 18m 37s | Posted - 2009.04.16 21:31:56
Quote: Originally Posted by lodenkamper Does LoopDeLoop give a multiple solutions message when multiple solutions are confirmed (i.e., solver finds 2 distinct solutions), or when multiple solutions are suspected (i.e., no further inferences possible)?
|
When running with the full solver, it gives the message only when it is confirmed. Internally with the iterative solver it will give it when the puzzle is incomplete but not proven impossible. In the UI it shows those as being incomplete, with the percentage of lines not filled in.
I'll try and find some time to redo it and capture both solutions to disk to verify that my program hasn't cheated somehow. Although the verification of correct solution code hasn't changed and is pretty thorough. |
Tilps Kwon-Tom Obsessive Puzzles: 6722 Best Total: 18m 37s | Posted - 2009.04.16 21:37:59
Quote: Originally Posted by pqg Tilips rated the difference in difficulty of his set pretty accurately (I'd rate the first one as comparable to a slightly above-average difficulty Naivoj beast, while the last one is more like a slightly easier-than-average BotM) Might I suggest that you need a better measure of how much the loops fills the area than proportion of vertices visited? As you've already noted, that often leads to large areas of 0s but what you may not have picked up on is that it also leads to the non-zeroed area being filled very densely by often having big areas of 2's which end up being all parallel lines.
|
I am aware of this problem, the 'loop expander' is just there to ensure the puzzle is at least not almost all 0's which is what used to happen all the time. The loop expander was a quick effort, so it only expands loops by changing 1 to "num edges of cell"-1 (my problems are always complicated by the fact I support arbitrary meshes), and can quite easily fall into the trap of creating very long segments. I'll try to adjust its random choice algorithm a bit to increase branching percentages against continuing parallell lines (whatever that means when extrapolated to arbitrary mesh...)
(Edit: Beta 39 has some minor improvements on this front, but I really need to do a fair bit more work on the loop generator before I will be happy with it.)
Last edited by Tilps - 2009.04.18 15:56:20 |
Tilps Kwon-Tom Obsessive Puzzles: 6722 Best Total: 18m 37s | Posted - 2009.04.18 15:52:31
Quote: Originally Posted by tilps I'll try and find some time to redo it and capture both solutions to disk to verify that my program hasn't cheated somehow. Although the verification of correct solution code hasn't changed and is pretty thorough.
|
Turns out my solver was broken, was a fairly subtle bug which was introduced when I added edge pair restriction support. The bug affects the iterative solver, but only causes the full solver to produce incorrect results.
I've fixed this in beta 39 and will see how long it takes to solve 99 now.
(Edit: 7minutes to solve 99)
Last edited by Tilps - 2009.04.18 15:58:06 |
v_e_e_n_c_a Kwon-Tom Obsessive Puzzles: 2080 Best Total: 32m 53s | Posted - 2009.04.20 12:32:53 I introduced coloring into my solver/generator. I will provide some beasts in future if you want - puzzles should be harder. With this feature it is now able to solve LDL6 in 0,086 seconds (it still doesn't use backtrack)
Last edited by v_e_e_n_c_a - 2009.04.20 12:33:53 |
lodenkamper Kwon-Tom Fan Puzzles: 21 Best Total: 47m 58s | Posted - 2009.04.24 05:39:08 I've just added some more user beasts in the beast thread.
The main questions I have for the folks that solve these are:
1) Are there puzzles among 100-114 and 127-136 that seem relatively much easier to solve than the others because of the availability of highlander deductions?
2) Where do 100-114 and 127-136 tend to lie on the difficulty scale set by Naivoj's beasts 1-99?
3) Do any of 100-114 and 127-136 seem markedly more difficult then the others? |
pqg Kwon-Tom Obsessive Puzzles: 6385 Best Total: 15m 37s | Posted - 2009.04.29 18:44:52 TBH, it was too long ago when I did 100-114, so I can't remember which of them were easier/tougher but I seem to remember them being fairly consistent. Of 127-136, I thought 130 was the most difficult, closely followed by 129 & 131. I found those 3 to be considerably tougher than the rest. The easiest was 136, but that might have been because I'd had the practice of the first 9 by then...
In terms of comparison with Naivoj's beasts, I'd say the average difficulty of your's is very slightly higher. What's also interesting to note is that both your's and Naivoj's have a distinct style which you can begin to recognise with repeated puzzles. (I think Mondsemmel also commented somewhere that Naivoj's have a certain flow to them that's quite unlike the BOTMs) It's somewhat like doing a cryptic crossword or pub quiz by the same setter regularly - familiarity with the style can point you in the direction of more useful enquiries sooner.
I used highlander techniques extensively in all of them - so much so that I would not relish attempting any of them without highlander.
Last edited by pqg - 2009.04.29 18:46:26 |
v_e_e_n_c_a Kwon-Tom Obsessive Puzzles: 2080 Best Total: 32m 53s | Posted - 2009.06.08 08:00:04 Please let me know about puzzle difficulty.. I still doesn't use any backtrack in my generator (no trial), but I added some features and I want to know how much difficult (or easy) they are.. Times for generating them - 25-30 sec.. Thanks.. |