| Discussions on User Beast Puzzles | Naivoj
Kwon-Tom Addict
Puzzles: 314
Best Total: 33m 50s | Posted - 2008.02.12 02:54:22
Quote: Originally Posted by chairman Should not the middle line of blue characters be o o o e ? edit: No it should not. Apologies. |
I have to say I am not following very well, in part because the drawing is not clear enough. Like the red e is too centered, so I can not tell if it is referring to inbound left bottom corner of clue 2 @c7r5 or to the inbound right bottom corner of the square @c6r5? Same thing with the first e on the first row ?
On my own I only get: NaivojPoorCornerFlagging.PNG
I do not understand how you can determine the corners I have marked with question marks ? | Jankonyex
Kwon-Tom Obsessive
Puzzles: 5680
Best Total: 9m 35s | Posted - 2008.02.12 06:36:46
 | Jankonyex
Kwon-Tom Obsessive
Puzzles: 5680
Best Total: 9m 35s | Posted - 2008.02.12 06:39:08
just external to internal, neither complicated nor confusing. | Naivoj
Kwon-Tom Addict
Puzzles: 314
Best Total: 33m 50s | Posted - 2008.02.12 16:33:54
I really want to figure out this; I suppose I need a legend, so I will start one with what I know, and I am asking anybody who understands more to complete it.
e stands for EVEN; An even corner has 2 lines or 0 lines(2 crosses), because 2 and 0 are even numbers.
o stands for ODD; And odd corner has 1 line (and 1 cross), because 1 is an odd number.
The diagonally opposite corner (through the nearby dot, so within the diagonally opposite square) of a corner has the same e/o attribute value;
What the blue lines mean?
- In the latest diagram some blue lines are in diagonal giving a cool 3 D effect (although I doubt that is the intention).
The gray lines (in the latest diagrams) seem to indicate how the loop-the-loop line is passing through a corner.
What the circles mean in the latest diagrams?
- probably it is just circling each corner with its associated gray lines
If somebody can explain how the first e in the first row was determined, there is a chance I will understand the rest.
- also is that e on the top right corner of a 2 clue, or on the top left corner of the blank square at the right of that 2? or on both corners? | Jankonyex
Kwon-Tom Obsessive
Puzzles: 5680
Best Total: 9m 35s | Posted - 2008.02.12 17:04:14
The two dots in the middle may not necessary to be considered as its surroundings is totally lain inside the blue loop and they must be neutral points n or simply e. | MondSemmel
Kwon-Tom Obsessive
Puzzles: 6159
Best Total: 7m 47s | Posted - 2008.02.22 16:17:47
I don't really get those colour rules (at all), but obviously all of them can basically be substituted by more trial & error. If I knew where to start, I'd probably try to learn them, especially for things like (most recent example) the top right part of puzzle #14 (a mass of 10 horizontally or vertically connected 2s - the puzzle was already nearly completed at that point, but without trial & error, I had no idea how to continue.
I'm grateful for those big puzzles, although I guess they're almost too hard for me  - I'd probably have to learn more patterns and colouring rules to solve them without the huge portions of trial & error in between.
As a status report, I've loaded the first 15 puzzles and then basically done the trivial parts in all of them in a row and then tried to finish them in a row - while accidentally deleting #11 in mid-progress. Among the first 15 puzzles, I've finished all of them except for #5, #11 and #15.
Seeing your difficulty ratings really puts things like personal strengths and weaknesses in solving these puzzles in perspective - you ranked #14 as easy, but I still needed masses of trial & error for one part. One the other hand, perhaps colouring rules could have helped me solve parts like that without much trial & error?
Thanks again for all those big puzzles, and keep them coming! (If you could show me how to generate more of them myself, I could try to help generate them, too) | Naivoj
Kwon-Tom Addict
Puzzles: 314
Best Total: 33m 50s | Posted - 2008.02.24 09:49:36
Quote: Originally Posted by mondsemmel I don't ... too) |
I believe the beasts I have posted, except the ones marked HARDER, are of the same level of difficulty than the Kwon-Tom Loop ones. Do you found them harder? Like the ones I have tagged as EASY are comparable (as per program stats) to the easiest KT beast posted. But I guess none of the KT beast are really easy either.
I am surprised that you completed #12 (supposedly the hardest as per program stats) before 5, 11 and 15. How tough was solving #12 ?
- Maybe you are making more highlander type deductions than the program.
- Will you attempt to solve "insane" #18?
- I am not sure if I say that before but @Janko beast #100 only ranks 8 among the 34 KT beasts posted (again as per program stats);
I suppose you (a club 19 member) are exaggerating when you are saying "I don't really get those color rules (at all)". Did you understood my color rules explanations for #2 Chairman position (on page 1 of current topic)? If not, do you understands the basic of coloring inside path one color and outside paths another?
I think the key deductions in solving #14 are the path ones (=closing loop). If you save a #14 position (without t&e) in a loopy file and post it (with RapidShare maybe), I'll be glad to point out forced moves, like I did for the puzzle #2 position Chairman posted. | MondSemmel
Kwon-Tom Obsessive
Puzzles: 6159
Best Total: 7m 47s | Posted - 2008.02.29 20:26:48
I've now done all 26 puzzles (I might possibly have destroyed my progress in #25 or #26, but I'm not even sure of that, and in any case, I've done the rest of them) as well as BotM 02/2008 as a comparison. If I had to rank it, I'd never put it as the hardest BotM there is ... when I tried solving it the first time, I made a mistake in the last quarter of the puzzle (and as I was doing it on paper and not with a pencil, I had to start over), it seemed to me that most parts of the puzzle seemed to come "automatically", and if it was really necessary to t&e between two possible paths, most of the tries immediately showed that only one was valid.
On the other hand, to give examples of two other BotMs, there were instances when the last puzzle of the puzzle seemed to be an entire puzzle by itself, with no more easy patterns, requiring much trial & error because everything seemed to work until the puzzle was filled ...
And in another instance, there was a puzzle where there were no easy ways to find out the solution for the top half; several cases of t&e brought meager results at best - and then I tried one at another place and the whole puzzle was solved without needing anything beyond trivial solution methods (This is always problematic, of course, because t&e is usually more rewarding if it immediately produces a logical mistake - then you know the other path is correct. If there is no logical mistake, you'll only know that the option you tried is correct if the puzzle is finished...).
Rating these puzzles really does seem incredibly hard. Of course, it's easy to say whether a puzzle is really easy or really hard, but anything in between can only be an approximation. As I said above, I've now solved all 26 of them, including #18. And while that puzzle was hard, there's no way that it was nearly four times as hard as the second hardest puzzle you posted. Maybe implementing advanced highlander deductions into the solver would significantly reduce solving times for this particular puzzle?
Tilps' suggestion of using virtual numbers sounds to me as if it would make detecting highlander deductions easier for the computer. As an example, take the top left 2x2 square of this example which consists of 4 empty fields:
In the top left corner of this 2x2 square there are only two numbers which yield unique solutions for this puzzle, whereas two others will yield the same solution and therefore counter each other out:
In the meantime, the numbers 0 and 2 (and 4, to be exact) will yield unique solutions:
If you put the valid solutions for 0 and 2 on top of each other, two crosses will exist in both solutions and therefore be the valid highlander deduction. I don't know whether this approach to highlander deductions is easy or useful enough to make it worthwhile to implement it, but maybe it's worth a try?
My memory of Janko #100 made it harder than it was. At the time I didn't yet know kwontomloop.com and I was much slower than I'm now. With the me then, #100 was definitely an incredibly hard puzzle (although your rankings, as I mentioned with that of #18, aren't perfect, so for now I'll just say that it's "better" than "a meager #8" ). Oh, and besides, don't forget that the standard java applet of Janko.at reduces the size of the biggest puzzles so much that the numbers are hardly recognizeable which does have an impact on solving speed. On the other hand, I still love the fact that you can click close to a corner in that applet and make two lines with one click ...
What I meant when I said that I don't understand the colouring rules is simply that I haven't yet seen a normal "introduction" to them - they aren't inherent to my solving style and therefore don't "fit into my thinking". And you have to admit that pictures like the one on this page that have random colours, letters and circles on them are nothing but confusing until you understand at least the rudimentary concepts. In addition, I'm not a native speaker in English, and at times that does prove to be an additional barrier to understanding technical stuff.
As an example I'm sure you'll agree with, take Jankonyex quote in this thread (although it has nothing to do with colouring rules, I think - not that I'm sure about that, either^^):
http://www.kwontomloop.com/forum.php?a=topic&topic_id=308
I don't understand a single thing in that (admittedly very technical and mathematical) quote - and because it seems so incredibly alien, it doesn't make me want to understand it, either - because it either seems that it doesn't fit into my solving style or because I fear that I simply won't understand it, I don't know.
Right now, the only thing I know about colouring rules seems to be that you colour the outside of the puzzle in one colour and the inside of the puzzle in another colour - but I don't have any idea in which way this helps one solve the puzzles .
By the way, you don't actually have to be incredibly skilled at solving puzzles to get solving times like mine. I have an RTS background and my mousespeed is really fast, and that means that I don't need much time for the trivial parts, which sometimes means the whole puzzle, sometimes almost nothing, but in any case, if I'm fast at that, I have more time for other parts.
Oh, and of course, I've already done tons of these puzzles, e.g. all those in the archive (some more than once, especially when I tried to do the Wrongs (something I've long given up now^^), all bigger ones on janko.at (many more than once), all BotMs (many more than once), etc. But that doesn't mean that my solving methods are better than those of somebody who needs a bit more time than me.
Last edited by MondSemmel - 2008.02.29 20:29:10 | PuzzleLover
Kwon-Tom Obsessive
Puzzles: 1033
Best Total: 38m 17s | Posted - 2008.03.13 05:52:35
Thanks for more beasts, Naivoj. What does the "(rot)" in a puzzle description mean? | Naivoj
Kwon-Tom Addict
Puzzles: 314
Best Total: 33m 50s | Posted - 2008.03.14 01:18:36
Quote: Originally Posted by puzzlelover What does the "(rot)" in a puzzle description mean? |
| Naivoj
Kwon-Tom Addict
Puzzles: 314
Best Total: 33m 50s | Posted - 2008.05.27 08:44:10
Quote: Originally Posted by mondsemmel Is there a chance of you making some more of these? I'd definitely try
them out. So far I've done every beast you posted, half of them twice,
but it's problematic in that I pretty much somewhat "remember" puzzles I
already did, so solving beasts again isn't as enjoyable/interesting as
solving them the first time is. Besides, knowing which parts of a puzzle
are hard beforehand (because I already know the puzzle) makes the
solving process, although not the specific solution, somewhat predictable. |
I will like to reserve the "User Beast Puzzles- Loopy Strings & Flash Links" topic strictly for posting puzzles, so it is shorter to browse and it looks more like the "Kwon-Tom Loop User Puzzles" page. Any comments on user beasts should be post in the current topic.
- So Mondsemmel since I just quote your last note, about you edit it out of the ""User Beast Puzzles" topic?
Regarding my rating, I now agree with previous criticisms from Tilps and Mondsemmel, especially when Tips was saying that if you solve a puzzle from another corner it could be drastically easier or harder, to which I replied that I tested that last year without noticing significant difference. But I had only had tested a few 10x10 puzzles and now realize that was not enough to conclude anything. Lately I tested some user beasts and found huge rating differences in many puzzles when starting from the opposite corner.
I still think that the idea of rating by counting the # of “human like” explorations is good, but I need to refine it. My idea is to upgrade our program to solve a puzzle 4 times from each corner and then calculate a prorated rating. But I have a hard time founding time to work on this so it would probably take a while. | lodenkamper
Kwon-Tom Fan
Puzzles: 21
Best Total: 47m 58s | Posted - 2008.06.01 02:25:11
Thanks much for the user beasts. They've been marvelous test cases for my solver. Doing them with pencil and paper will also keep me out of trouble for a long time. A few notes and observations:
1) Timing for my solver for #18 is ~20-40 s, depending on solver parameters that affect (indirectly) how the searching is done. This is on a 2.8 GHz iMac. About 5-6 of the user beasts take longer to solve than #18, with my solver.
2) I agree that rating puzzle difficulty is not easy. The most interesting case for my solver is #41, which solves in about 90s or not at all (i.e., solver time > 1 hour) depending on solver parameters. I think it's possible for a search to effectively get stuck, so luck starts to affect computer solving times.
3) There appear to be multiple solutions to user beast #5. All others to date appear to have unique solutions. I have PDF files of two distinct solution, if there is some easy way to post them. | MondSemmel
Kwon-Tom Obsessive
Puzzles: 6159
Best Total: 7m 47s | Posted - 2008.06.01 16:34:28
I wanted to say thanks once again for the beasts. I've now solved all the new ones up to this point (#56) except for #49, in which I somehow managed to make the same mistake two times in a row (i.e. after restarting, I somehow ended up in the same wrong situation again) where the puzzle is diagonally seperated into two parts (meaning the result will consist of at least two loops). I'll probably try to solve it sooner or later again, but seeing the solution once would help, too.
Puzzle #18 has several cases of a bit more complex highlander deductions, I think. I also didn't think it was the most difficult puzzle, but it definitely was among the more difficult ones.
However, I found (nearly) all of Naivoj's puzzles harder than all the Beasts of the Month. I recently did all of those, too, and with a few exceptions, I had solving times of ~25-30 minutes for each of them.
I can't really talk about ratings as I don't have a solver to solve these puzzles, though (and besides, it would probably tempt me to look at the solution every time I'm stuck at a hard part).
And again, thanks a lot for those puzzles! I really enjoy solving them  .
I'm really interested in how you generate these puzzles - could you elaborate a bit on that, please? (I especially mean at this difficulty - I use Loopy to solve puzzles, but the ones that Loopy generates are trivially solvable for it, too, and it's the same for me - basically nothing more than lots of brainless clicking. In contrast, all puzzles here are a challenge and actually require thinking^^). | Naivoj
Kwon-Tom Addict
Puzzles: 314
Best Total: 33m 50s | Posted - 2008.06.02 09:27:45
Quote: Originally Posted by lodenkamper 1) Timing for my solver for #18 is ~20-40 s, ... This is on a 2.8 GHz iMac ...
2)... The most interesting case for my solver is #41, which solves in about 90s...
3) There appear to be multiple solutions to user beast #5. ... I have PDF files of two distinct solution, if there is some easy way to post them. |
But when solving it from the opposite corner:
- The no of "human like" explorations is reduced by 2.4 times
- The CPU time go down to 4 min 12 sec (Edit: 2m43s now)
You definitely have developed a fast solver for the beasts!
2) For me #41 requires 1 min 17 sec. But I just solve it from the opposite corner for the very first time,
and it is twice as hard for our solver:
- The no of "human like" explorations is multiplied by 2
- The CPU time is multiplied by 2.2 for 2min 51sec
3) Tilps program LoopDeloop agrees with you, so this would indicate a bug with our program. About using RapidShare to post your solutions here. Note that RapidShare has been used in the current topic to post puzzle positions. Btw I would prefer loopy save files then PDF files.
MondSemmel, I would reply to your note in the next few days.
Last edited by Naivoj - 2008.06.11 10:35:43 | Naivoj
Kwon-Tom Addict
Puzzles: 314
Best Total: 33m 50s | Posted - 2008.06.06 06:59:31
Quote: Originally Posted by mondsemmel I wanted ... except for #49, ...but seeing the solution once would help, too.
Puzzle #18 ... I also didn't think it was the most difficult ...
...as I don't have a solver ...
I'm really interested in how you generate these puzzles - could you elaborate a bit on that, please? (I especially mean at this difficulty)... |
- Yeah I found out that #18 is 2+ times easier for our solver when starting from the opposite corner.
- But I will be posting beast #73 sometime soon that will be without a doubt the beastier.
- I think only programs should try to solve it, but I am sure you will try.
- I need to make totally sure of the uniqueness first.
- If you are looking for a public solver program, Tilps' enhanced Loop-De-Loop 3 Beta 36 is now an excellent beast solver
- It has solved most of the beasts I have tried, except #63, #69 and upcoming #73.
Our puzzle generator I think works like most, except maybe in the precision in which you can obtain a requested difficulty. I have already given some details in the "puzzle rating systems?" topic, which you should read if you haven't. I should give more details, but first I'll ask permission to the main programmer of this solver/generator.
Last edited by Naivoj - 2008.06.08 06:45:59 | lodenkamper
Kwon-Tom Fan
Puzzles: 21
Best Total: 47m 58s | Posted - 2008.06.06 23:11:42
Quote: Originally Posted by naivoj 3) Tilps program LoopDeloop agrees with you, so this would indicate a bug with our program. About using RapidShare to post your solutions here. Note that RapidShare has been used in the current topic to post puzzle positions. Btw I would prefer loopy save files then PDF files. |
| Naivoj
Kwon-Tom Addict
Puzzles: 314
Best Total: 33m 50s | Posted - 2008.06.07 00:20:50
Quote: Originally Posted by lodenkamper Please let me know if it would be helpful for you to see the solutions to #5 that I've found. If these are of interest, I'll see about getting them into loopy save file format. |
Quote: Originally Posted by lodenkamper
3) There appear to be multiple solutions to user beast #5. All others to date appear to have unique solutions.
|
Could you please test #57-65? | lodenkamper
Kwon-Tom Fan
Puzzles: 21
Best Total: 47m 58s | Posted - 2008.06.07 04:38:53
Quote: Originally Posted by naivoj
Could you please confirm you tested user beasts #01-56 with your program solver and found out that only #5 as not unique?
Could you please test #57-65?
|
Thanks again for posting the user beasts. | Naivoj
Kwon-Tom Addict
Puzzles: 314
Best Total: 33m 50s | Posted - 2008.06.08 06:38:24
Quote: Originally Posted by lodenkamper I can confirm that user beasts #1-4, and #6-65 all have unique solutions according to my solver. |
Could you please do more validations?
- #5, as I corrected it: Loop-De-Loop 3 Beta 36 is now solving it which means it is unique.
- And my newest #66-69 (Loop-De-Loop could not solve #69 after 8 hours)
Stay tune for upcoming #73 which will be by far the beastier posted! | lodenkamper
Kwon-Tom Fan
Puzzles: 21
Best Total: 47m 58s | Posted - 2008.06.08 20:23:28
The revised #5, and #66-69 have unique solutions according to my solver. |
|
