Sudoku Explainer: Bugs, Quirks and Other Remarks...

Programs which generate, solve, and analyze Sudoku puzzles

Sudoku Explainer: Bugs, Quirks and Other Remarks...

Postby 999_Springs » Fri Aug 27, 2010 8:06 pm

[Edit: There was supposed to be some message at the top here but I think it got edited out by moderator. WHAT!!?!?]

1. SE is really, REALLY BAD at spelling. In their XY(Z)-wing technique it talks about "occurances". Now that is THE WORST spelling error of "occurrences" that I have ever seen. It should hang its head in shame. (If it has one.) (See #73.)

2. SE's four options for the "Look & Feel" are UNBELIEVABLY BAD. Even I could design better graphics than that. And I know a bunch of others who could do much, much better...

3. SE doesn't recognise "UR/UL Type 3 with Hidden Pair" if either of the two cells at the intersection of the UR/UL and the subset contain 5 or more candidates. That is SO STUPID. Type 3 with Hidden Pair only relies on two trilocations, regardless of numbers of candidates in each cell.

4. What the heck was going on with all those 6.699999999999999 ratings of earlier versions? I'm not really a physicist myself, but I'll tell you that there are way too many SIGNIFICANT FIGURES in this rating. I would lose marks for doing stuff like that in my exams.

5. SE is REALLY SLOW on Mondays. One Monday I tried to rate this 6.5 puzzle and it took almost 6 minutes! Then I tried to rate a 10.2 and that took 19 MINUTES! What is wrong with it!? Well, the truth is that I haven't tested it out on other days of the week, but I fear that it may not be faster than this on Fridays.

6. Or Sundays, or Thursdays, or Tuesdays, or Saturdays, or Wednesdays.

7. My copy of SE doesn't work underwater. How primitive is that technology!? What if I need to rate this puzzle that I find at the bottom of a shipwreck? How do I do that?

8. Apparently m_b_metcalf has managed to attach a cattle prod to his copy of SE. I wonder how he does that.

9. How do you rate puzzles in batch using SE on Windows XP?

10. I don't know how the "Load" function under File actually works. It came up with a list of .txt files on my computer, of which I don't have many, so I put in all that I could find.
Firstly I tried a file called "BLUR Round 2 Demos", and I got this:
Code: Select all
.........
.........
.........
118457358
.........
112565676
111459782
111498152
114487364

Now that looks nothing like the original file or a Sudoku. What!?
Then I tried "BLUR Round 3 Demos":
Code: Select all
.........
116459475
.........
114414671
114178864
113278484
113675992
116457896
.........

Then I tried "IGCSE English Language Reading Paper 2005" and got this:
Unreadable Sudoku format

(Well at least it has SOME common sense!)
Then I tried "Part 1 - Basic Teleporter" and got this:
Code: Select all
1........
1.....1..
.........
.66......
......4.5
4........
.........
6........
.........

Enough said.

11. SE has a bug in their BUG that really bugs me. (Excuse the pun.) (Well, it's not a bug as such, but...) In their BUG Type 1, the program eliminates the two values in the correct cell, but it doesn't assign the third value to that cell, which it should. Since SE finds hidden singles before naked singles, the BUG cell is left as a blank cell with one candidate in it until it gets wiped by hidden singles. That's just stupid. Nobody solves like that.

12. Direct Claiming doesn't exist, except possibly in the first two fake "BLUR competition demos puzzles" mentioned beforehand. Well, then, why include it as a solving technique!?

13. SE rates "Bidirectional X-Cycle (6 nodes)" as 6.5 but a Turbot Fish as 6.6. What is that Bidirectional X-Cycle? It is an Unfinned Mutant Swordfish. Now WHY IS THAT FOUND BEFORE A TURBOT FISH!? What is WRONG with SE!? Does it really despise fins that much to give any sort of finned fish such an inflated rating!?

14. Has anyone found a non-degenerate, standard, normal Forcing Chain rated 7.6 or 7.7? Degenerate ones do exist, but I'm thinking that they're pointless...

15. Stupid bug here:
Put a valid puzzle into SE. It must not be a diamond, but anything else is fine.
Click on "Check validity".
Click on "Get next hint".

16. That "Quit" button is completely pointless. I mean, COMPLETELY POINTLESS. So ridiculous.

17. Suppose you're analysing a puzzle rated between (say) 9.0 and 10.1 (in particular the higher end). You want it to stop what it is doing because you either hit F9 by accident or it's taking too long. How do you do that? You can't. You just have to wait. Seeing as this can take over 5 minutes to do this, your frustraton at the speed of your computer turns into anger, and your anger leads you to do unwise things to your computer using sharp objects...

18. ... and if your puzzle is rated higher than 10.1, it gives you that "may take a long time" message, but... HOW LONG!? It doesn't specify! And how can I stop it if it's taking too long? BAD COMPUTER!! HURRY UP!! AAAAAAAARGH!!!

19. Suppose you've just clicked on "Solve Step". And you hit F8 (to see the solution), and your thumb completely accidentally comes into contact with... spacebar. Where is the original puzzle!? Part way through solving it... and... ARGH! Where's the undo button?

20. Seriously, SE needs an undo button. If you're going through the pencilmark grid and accidentally set a cell, how do you get your original PMs back? SE resets them automatically on removal of a given cell. And there is the problem mentioned above, which is REALLY TEDIOUS if you accidentally hit F8+click on Solve step/Apply hint/Get next hint/Get all hints. That happens to me a lot... WHERE'S THE STUPID UNDO BUTTON!?

21. Suppose you're rating a puzzle that's only of modest difficulty, and you accidentally hit F9 twice instead of once... you're doomed.

22. I'll have to take issue with SE's treatment of low-clue puzzles. Suppose you put into SE a puzzle with hardly any clues, and hit F7, and it tells you that there aren't enough clues. Now get this: SE's grammar is ABSOLUTELY HORRIBLE. I mean, COME ON! What is wrong with it!? "Sudokus with 16 clues or less..." Everyone KNOWS that it should read "FEWER", not "LESS"!

23. And for one-clue puzzles, it reads "This Sudoku has only 1 clues." You mean 1 CLUE, without the S.

24. Unproven assumptions! Oh dear! "The minimum number of clues must be 17." Nobody knows that! There might be a 16 lurking out there! SE shouldn't jump to conclusions!

25. OK, I put into SE that 16-clue 2-solution puzzle and manually removed candidate 8r9c3 so that it had a unique solution... and hit F7. It says "not valid", and makes a note about there being 16 clues. WRONG!!!!!

26. And then I clicked "Get next hint"... and it makes a note about 8 and 9 not appearing in the grid, rendering the puzzle invalid. WRONG AGAIN!!!!! Then I clicked "Get next hint" again... and... it bugs as per #15.

27. SE doesn't recognise "UR/UL Type 3 with Hidden Triplet" if either of the two cells at the intersection of the UR/UL and the subset contain 6 or more candidates. That is SO STUPID. Type 3 with Hidden Triplet only relies on three quadrilocations, regardless of numbers of candidates in each cell.

28. Who would want to turn "High quality rendering" off, anyway!?

29. Why doesn't SE recognise W-Wing?

30. Why doesn't SE recognise M-Wing?

31. Now comes The Ultimate Holy Grail Of Stupidity. SE finds "UR Type 3 with Hidden/Naked Quad" BEFORE a simple Naked/Hidden Quad. Enough said.

32. SE's treatment of quads is even more inconsistent than that! It rates a Naked Quad before a Hidden Quad, but it rates a UR Type 3 with Hidden Quad ahead of a UR Type 3 with Naked Quad!

33. And continuing with the quads, SE doesn't recognise "UR/UL Type 3 with Hidden Quad" if either of the two cells at the intersection of the UR/UL and the subset contain 7 or more candidates. That is SO STUPID. Type 3 with Hidden Quad only relies on four quintilocations, regardless of numbers of candidates in each cell.

34. Still going with the quads, how come it is the case that "Direct Hidden Pair/Triplet" exists... but Direct Hidden Quad doesn't exist? (I verifed this by manual removal of pencilmarks.) Another glaring inconsistency!

35. Oh look, SE's own generator can generate stuff with a half-decent rating. I thought that it never did anything past 7.3, even though it pretends to go up to 10.0.
Code: Select all
6.. .5. ..4
.1. .9. .3.
... 6.3 ...

..6 8.7 2..
27. ... .96
..1 9.2 8..

... 1.4 ...
.3. .8. .6.
4.. .2. ..9 SE=8.3


36. Continuing the pattern of #3, #27 and #33, it would seem that if you have a UR with only one bilocation attached to it, it would count as a "Type 3 with Hidden Single", which certainly does make sense. However, you will know that this is more commonly known as "UR Type 4". Now why does SE find "Type 3 with Hidden Pair" before "Type 3 with Hidden Single (a.k.a. Type 4)"? That is just illogical!

37. SE might be the only solver I know of which does nested forcing chains, but there are a few issues to clear up. Now in a typical Contradiction chain with nested chains, you'll get something at the top which reads like:
With this solving technique, we will prove the two following assertions:
If R2C1 does not contain the value 1, then R1C2 must contain the value 8
If R2C1 does not contain the value 1, then R1C2 cannot contain the value 8

and you'll expect the first two chains to have headings like that. But no. There is a bug in its treatment of Contradiction chains which causes them to do this:
Chain 1: If R2C1 does not contain the value 1, then R1C2 cannot contain the value 8 (View 1):
Chain 2: If R1C2 must contain the value 8, then R1C2 cannot contain the value 8 (View 2):

Clearly incorrect descriptions. Needs to be fixed.

38. Similarly, if there is a Bidirectional Cycle nested inside a chain, SE finds the wrong words to describe it. Like this:
Nested Bidirectional Y-Cycle: R6C4,R8C4,R8C9,R2C9,R2C4
Chain 3: If R2C4 contains the value 3, then R2C4 must contain the value 3 (View 3):
Chain 4: If R2C4 does not contain the value 3, then R2C4 cannot contain the value 3 (View 4):

How stupid. Fix immediately!

39. SE doesn't find the simplest step at each case of the main chains at the top. Like here:
(6) If R1C3 contains the value 4, then R3C1 cannot contain the value 4 (the value can occur only once in the block)
(7) If R3C6 does not contain the value 7 (2), then R2C1 cannot contain the value 4 (Forcing Chain: R2C1.4 off)

Any idiot could work out for what OTHER reason r2c1 can't be 4 just by reading this.

40. SE doesn't give the same rating to isomorphic puzzles. Like these two:
Code: Select all
 . . 3 | 4 5 6 | 7 . .
 . . 9 | . . . | 1 . .
 . . 5 | . . . | 2 . .
 - - - + - - - + - - -
 . 8 . | . . . | . 9 .
 5 . . | . 8 . | . . 7
 . 4 . | . . . | . 6 .
 - - - + - - - + - - -
 . . 8 | . . . | 4 . .
 . . . | 9 . 7 | . . .
 . . . | . 2 . | . . .   ER=8.9

 . . 3 | 4 5 6 | 7 . .
 . . 8 | . . . | 5 . .
 . . 9 | . . . | 1 . .
 - - - + - - - + - - -
 . 6 . | . . . | . 4 .
 3 . . | . 2 . | . . 5
 . 1 . | . . . | . 2 .
 - - - + - - - + - - -
 . . 4 | . . . | 2 . .
 . . . | 1 . 3 | . . .
 . . . | . 8 . | . . .   ER=8.5

Ha ha. Found on page 18 of patterns game 1.0.

41. As a corollary of #40, SE doesn't always find all the region forcing chains when they are there. See RW's reply to the isomorphic blunder on that thread.

42. Starting from the blank grid, manually remove candidate 1 from r2378c2378 and force SE to get the next elimination. It doesn't find the fish. It just lags.

43. SE has a BIG bug in its Unique Loop technique. If you put into SE a puzzle with six different numbers completed in all nine of their locations, and scant values of the other three numbers, then SE tries to find a Unique Loop with the remaining three values, even though none exists. At each step, it takes about a minute to verify that there isn't a Unique Loop in the grid. As a result, it is possible to create "lag puzzles" that are deliberately designed to stump SE even though their rating is less than 7.

44. SE doesn't sell winning lottery tickets.

45. SE doesn't provide surfing (waves, not internet) lessons.

46. SE's generator is useless at the Patterns Game.

47. SE doesn't let you rename the techniques. For example, if I wanted to rename a direct hidden triplet as a "You see, the thing about these instructions are that they are in three different languages - they start in English, but then they go into French and Spanish. How stupid - if they wanted people to understand them then they should write them all in English - so that we have a greater chance of being able to read all of it! Now we have to be trilingual just to read the stupid directions for this thing!", you find that you can't rename the technique.

48. The (+) in "Forcing Chains (+)" only finds locked candidates and pairs and X-wings. It doesn't find triplets. A triplet of ab-ac-bc type is found as a nested three-cell bidirectional Y-cycle with bug (as indicated by #38) and any other type of triple isn't found at all.

49. That other day I was using SE to guide me through a puzzle. And then I clicked on "Get next hint", expecting it to find the really obvious XY-Wing after the hidden pair. But - it just completely stopped responding, without any warning message or provocation. I had to load up another copy of SE and get it to the same position to continue with the puzzle. (Oh and by the way, what day of the week was it? #5 should give you a clue - it's described in most of the hint.)

50. I had my birthday recently. (WOOT!!!!) But when I opened up SE, it didn't give me a present. Only that really cheesy welcome message. UGH. Now I'll show it. When SE has its birthday I won't give it a present. Serves it right.

51. Aligned Triplet Exclusion doesn't find all aligned triplet exclusions. Take puzzle #19 from the top95:
*INSERT PUZZLE HERE*
Once it solves to the point of the Naked Single r6c2=9, there is an aligned triplet exclusion of r5c2, r6c2 and r6c3 that eliminates 1r6c3 and 4r6c3 using cells r6c1 and r3c2. There are also three others and I'll leave it as an exercise to you to find them. But SE just leaves them and goes to Nishio.

...

...

...

[October 10th]

52. I've just realised that I left the first 51 points on a file on my computer back in July. I actually lost interest in writing this thing! Therefore, SE is VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY VERY BORING!!!!

53. I wish SE had more free money to give away.

54. Direct Pointing is Directly Pointless. I mean, COME ON! No-one solves puzzles like THAT! If SE thinks that people look for Pointing only when it reveals a hidden single in the box, then it should think again. That's clearly nonsensical.

55. Here's SE's treatment of ULs:
Code: Select all
size      4    6    8   10   12   14   16
Type 1                                     12.3
Type 2                                     12.3
Type 3+NP                                  12.4
Type 3+HP                                  12.3
Type 3+NT                                  12.5
Type 3+HT                                  12.4
Type 3+NQ                                  12.6
Type 3+HQ                                  12.5
Type 4                                     12.3
Type 5                                     12.3
        -7.8 -7.7 -7.6 -7.3 -7.3 -7.3 -7.3

In order to find a certain SE rating for a UL, you take the row and the column that your UL is in, and add the corresponding values at the end of the column and row. So for a simple UR Type 1, the rating is 12.3 + (-7.8) = 4.5. Now in my point of view, that is just completely illogical. It just doesn't make any sense.

56. There's a 5.3 in the above table. Has anyone found a puzzle rated 5.3 yet? I don't know.

57. Oh yeah, someone should edit the above table to include all the ratings that end in 99999999999999 which I didn't include.

58. "Direct Hidden Pair"!? WHAT!? Whose idea was that? At best it's an April Fool's joke ( ... when was SE created? December or something?), and at worst it could be... something beyond the realms of our imagination. NOBODY SOLVES PUZZLES LIKE THAT. I know that there may be the odd occasion when a hidden pair is found before a naked single, but even in those ridiculously rare cases it's irrelevant whether or not the pair produces a hidden single in the same house. To be continued...

59. And some "direct hidden pairs" are really obscure. In other words, a manual player isn't going to find them without pencilmarks. Ever. As a result of this, some puzzles are REALLY underrated by SE. Take Bob Hanson's default puzzle. That's SE 2.0. It seems a lot harder than that, at least from my point of view, anyway.

60. When you're solving a puzzle, it is sometimes intuitive to actually look for naked singles in the puzzle. That's perfectly sensible. But you don't ever scan the puzzle for hidden pairs at the beginning and you don't discard them if you see that there's no resultant single. SE's solving methods are ridiculous.

61. In the worst cases, a direct hidden pair gives a single that doesn't do anything. Then SE finds the actual hidden pair directly after that, which completely wrecks the point of the hidden pair.

62. SE claims that a manual player would solve puzzles like this in terms of looking for direct hidden whatevers and the like. I know when I'm being brainwashed by SE's conspiracies.

63. Even worse are their Direct Hidden Triplets. Nobody would EVER look for direct hidden triplets without pencilmarks. EVER. Not even RW. I'd be surprised if he ever found some that way. To be honest, I don't know why these were included as a technique in the first place.

64. Their generator for "hard"-level puzzles goes up to SE 2.5 but doesn't list direct hidden triplets as part of their description. That's just a blatant cover-up.

65. Nobody finds Direct Hidden Triplets before Pointing/Claiming. BLEH.

66. SE's ratings up to about 6.5 are extremely arbitrary and have no meaning. That is unbelievable. As a result of this, mean ratings carry less significance than they should, since the ratings are subjective. In particular their direct hidden whatevers.

67. A couple of months back, denis_berthier asked why SE 9.3 puzzles were so much rarer than SE 9.0-9.2 puzzles. I think I have the answer. Or at least some of it.
Consider puzzle #2 from the top1465:
Code: Select all
7.8|...|3..
...|2.1|...
5..|...|...
---+---+---
.4.|...|.26
3..|.8.|...
...|1..|.9.
---+---+---
.9.|6..|..4
...|.7.|5..
...|...|...

Now put it into SE and hit F9... the rating should come out as 9.5. Now I'm claiming that this is a conspiracy. Why, I hear you ask? It's because the actual SE rating is not even close to the generated value of 9.5, but is actually the elusive value of 9.3. How is this? Well, first you go through the puzzle and do all the easy steps. Then click on "get next hint" until you reach the "Dynamic Whatever Chains (+)" to eliminate 7r5c2. That is a short chain involving pairs inside it which is rated 9.3. Now hit the apply hint button, and hit F9 again. It should now be rated 9.2. So the puzzle can be solved with moves of SE 9.3 or less. Why does SE rate it 9.5? That's because if you "solve step" through the puzzle, SE gives all the Dynamic Whatever Chains without the (+) before moving on. Some of these can go up to SE 9.6, but they are found first before the dynamic (+) chains of rating 9.3. In other words, SE picks higher-rated stuff before the 9.3 steps, which push up the rating. That is a conspiracy. SE should always find the lowest-rated step at EVERY stage, and not give us this horrible rubbish involving SE 9.6 chains that nobody will ever find and, undoubtedly, they're aesthetically displeasing anyway, compared to the much shorter 9.3 chains.

68. I wish SE could invert its colours.

69. Actually, in addition to #65, no-one EVER looks for Direct Hidden Triplets - and even when you're daft enough to try to look for them, there's no way you can find them. Now SE claims that a "direct hidden triplet" doesn't need pencilmarks to spot. That is ridiculous. What were they thinking when they designed SE!? I bet you that even RW can't find these things without difficulty. And rating them 2.5 is just ridiculous.

70. Adding SE ratings together makes no sense at all. As mentioned in #66, the ratings are just arbitrary - and adding them up!? Well, that makes three hidden pairs in a puzzle look like the same difficulty as a 10.2 chain. Nonsense.

71. Same as #61 but with triplets. Bleh.

72. SE got 0 out of 40 on my "nearly impossible quiz". The lowest mark that anyone else has ever got on it is 11. Therefore SE is stupid.

73. This is HILARIOUS. You have got to see this. Re-read #1 again. Now look at this website: http://www.urbandictionary.com/define.php?term=occurances
SE is now officially classified as moronic! HA HA HA HA HA HA HA! (Note: I was completely unaware of this when I wrote #1.)

74. When you put an empty grid into SE, remove a big bunch of candidates, and click "get next hint" a few times ignoring those dumb messages about multiple solutions etc., you'll find an error message which says "java.lang.illegalArgumentException: countBits > 64". Um, that's definitely an unresolved bug. I have no idea what that means.

75. Actually, in addition to #46, I conclude that SE's generator is useless at EVERYTHING.

76. When SE 1.2.1 was created, the ratings for SE 11+ puzzles had been changed and they all had to be re-rated. Well. That's how inconsistent SE is. It has to change EVERYTHING. Why can't it just keep the ratings the same!?

77. Why is a UL 10 rated 0.3 more than a UL 8? (See #55.) It should only be worth 0.1 more. And a UL 16 is rated the same as a UL 10. That's illogical.

78. GRRR I HAVE STUPID ANNOYING PHYSICS COURSEWORK FOR NEXT MONDAY AND THIS DUMB SUDOKU EXPLAINER PROGRAM IS OF ABSOLUTELY NO USE WHATSOEVER. GRRRRRRRRRRRRRRRR.

79. On loading SE the box on the right of the screen says "Hints (0 hint)". Its grammar is appalling.

80. SUDOKU EXPLAINER IS ONLY ABOUT 90 TO 95 PERCENT RESPONSIVE TO MOUSE CLICKS. THAT IS SO *!@#-ING STUPID AND ANNOYING, ESPECIALLY WHEN YOU'RE TRYING TO DO STUFF QUICKLY. EVERY PROGRAM THAT YOU AN DOWNLOAD OFF THE INTERNET MUST AT LEAST BE CAPABLE OF RESPONDING CONSISTENTLY TO THE MOUSE. IT SEEMS THAT SE DOESN'T HAVE THIS CAPABILITY. HOW RIDICULOUS. *!@# THAT.
Last edited by 999_Springs on Sat Oct 09, 2010 8:37 pm, edited 1 time in total.
Hidden Text: Show
Once upon a time I was a teenager who was active on here 2007-2011
999_Springs
 
Posts: 367
Joined: 27 January 2007
Location: In the toilet, flushing down springs, one by one.

re: Sudoku Explainer: Bugs, Quirks

Postby Pat » Tue Sep 21, 2010 9:58 am

Pat (2008.Mar.9) wrote:

        what's needed out of all that's available
    the report from SuDoku Explainer
    usually lists all that's available ( including red herrings )
    -- or so i thought
then came this example,
where SuDoku Explainer reports only 2 Swordfish ( those 2 which are needed ! )
and neglects to mention 3 other Swordfish ( red herrings )


Mauricio (2008.Feb.26) wrote:

      [ first puzzle ]
Code: Select all
..1..2..3.2..1..4.5..6..1....2..3..7.7..4..3.6..8..4....5..6..9.6..8..7.8..5..3..


Code: Select all
 . . 1 | . . 2 | . . 3
 . 2 . | . 1 . | . 4 .
 5 . . | 6 . . | 1 . .
-------+-------+------
 . . 2 | . . 3 | . . 7
 . 7 . | . 4 . | . 3 .
 6 . . | 8 . . | 4 . .
-------+-------+------
 . . 5 | . . 6 | . . 9
 . 6 . | . 8 . | . 7 .
 8 . . | 5 . . | 3 . .

User avatar
Pat
 
Posts: 3440
Joined: 18 July 2005

re: Sudoku Explainer: Bugs, Quirks

Postby Pat » Tue Sep 21, 2010 10:06 am

Pat (2008.Jan.20) wrote:

        "Direct Pointing" and "Direct Claiming"
example B ( same as example A plus 2 more known cells r2c9 = r4c7 = 5 )

Code: Select all
 1 9 8 | 5 6 3 | 4 7 2
 . 4 2 | 9 . 8 | . . 5
 . . . | 4 . 2 | . 9 8
-------+-------+------
 2 3 9 | 7 8 6 | 5 . .
 8 . . | . . . | 2 . 9
 . . . | 2 . 9 | 8 . 7
-------+-------+------
 4 2 . | 8 . . | 9 5 6
 6 8 . | . 9 . | . 2 .
 9 . . | 6 2 . | . 8 .



the 4 for b9 is in c9, immediately creating a "hidden single" 4 in r4 -- the "hidden single" is created in a line -- not in a box -- and SuDoku Explainer considers it "Pointing" -- failing to recognize it as "Direct Pointing" !!


example C ( same as example A plus 2 more known cells r2c7 = 6, r4c8 = 4 )

Code: Select all
 1 9 8 | 5 6 3 | 4 7 2
 . 4 2 | 9 . 8 | 6 . .
 . . . | 4 . 2 | . 9 8
-------+-------+------
 2 3 9 | 7 8 6 | . 4 .
 8 . . | . . . | 2 . 9
 . . . | 2 . 9 | 8 . 7
-------+-------+------
 4 2 . | 8 . . | 9 5 6
 6 8 . | . 9 . | . 2 .
 9 . . | 6 2 . | . 8 .



the 1 for r4 is in b6, immediately creating a "hidden single" 1 in c8 -- which in my view qualifies as "Direct Claiming" -- but SuDoku Explainer considers it "Claiming" -- failing to recognize it as "Direct Claiming" !!


    i'd say it's just one bug
    which recognizes "Direct Pointing" and "Direct Claiming" only when the immediate "hidden single" is created in a box,
    failing to recognize it when it is created in a line.
User avatar
Pat
 
Posts: 3440
Joined: 18 July 2005

puzzle rated 4.6 can have isomorphs rated 6.6 7.1 7.2

Postby Pat » Thu Oct 07, 2010 2:22 pm

999_Springs wrote:
40. SE doesn't give the same rating to isomorphic puzzles. Like these two:
Code: Select all
 . . 3 | 4 5 6 | 7 . .
 . . 9 | . . . | 1 . .
 . . 5 | . . . | 2 . .
 - - - + - - - + - - -
 . 8 . | . . . | . 9 .
 5 . . | . 8 . | . . 7
 . 4 . | . . . | . 6 .
 - - - + - - - + - - -
 . . 8 | . . . | 4 . .
 . . . | 9 . 7 | . . .
 . . . | . 2 . | . . .   ER=8.9

 . . 3 | 4 5 6 | 7 . .
 . . 8 | . . . | 5 . .
 . . 9 | . . . | 1 . .
 - - - + - - - + - - -
 . 6 . | . . . | . 4 .
 3 . . | . 2 . | . . 5
 . 1 . | . . . | . 2 .
 - - - + - - - + - - -
 . . 4 | . . . | 2 . .
 . . . | 1 . 3 | . . .
 . . . | . 8 . | . . .   ER=8.5



it was reported by JPF (2007.Aug.28)

more recently, in a discussion now lost,
RW (2010.Mar.10) showed examples where
a puzzle rated 4.6
can have isomorphs rated
  • 6.6
  • 7.1
  • 7.2
User avatar
Pat
 
Posts: 3440
Joined: 18 July 2005

Re: puzzle rated 4.6 can have isomorphs rated 6.6 7.1 7.2

Postby ronk » Fri Oct 08, 2010 3:19 pm

Pat wrote:in a discussion now lost,
RW (2010.Mar.10) showed examples where
a puzzle rated 4.6
can have isomorphs rated
  • 6.6
  • 7.1
  • 7.2

I found these in the Patterns Game.

Code: Select all
original:
 1 . . . . . . . .
 . . 2 1 . . 3 . .
 . 3 . . 4 . . 2 .
 . 4 . 3 . 5 . . .
 . . 5 . 2 . 6 . .
 . . . 6 . . . 7 .
 . 2 . . 7 . . 8 .
 . . 3 . . 8 9 . .
 . . . . . . . . 4    ED=4.6/2.5/2.5

morph:
 . 8 . . 9 . 3 . .
 . . . 4 . . . . .
 . . 7 . . 8 . 2 .
 . . 4 . . 2 . 3 .
 . . . . . . . . 1
 1 . . . 3 . 2 . .
 3 5 . . . . . 4 .
 . . 2 . 6 . 5 . .
 6 . . . . 7 . . .    ED=6.6/2.5/2.5

Code: Select all
original:
 1 . . . . 2 . 3 .
 . . 2 4 . . 5 . .
 . 6 . . 5 . . . 7
 . 1 . . . . . 8 .
 . . 6 . 3 . . . 2
 5 . . . . . 6 . .
 . 8 . . . 4 9 . .
 7 . . 5 . . . . 8
 . . 3 . 9 . . 4 .    ED=4.6/2.0/2.0

morphs:
 . . . 5 . . 6 . .
 3 . . . . 6 . 2 .
 . . . . 1 . . . 8
 5 . . . 6 . . 7 .
 . 2 . 1 . . . . 3
 . . 4 . . 2 5 . .
 . . 5 7 . . . 8 .
 9 . . . . 3 . . 4
 . 4 . . 8 . 9 . .    ED=7.1/2.0/2.0


 . . . 2 . . . 4 .
 . . . . . 4 5 . .
 . . 3 . 5 . . . 9
 . . 6 . . 2 . . 3
 1 . . . 6 . . 8 .
 . 5 . 1 . . 7 . .
 8 . . 3 . . . . 4
 . . 2 . 7 . 8 . .
 . 6 . . . 5 . 9 .    ED=7.2/2.0/2.0
ronk
2012 Supporter
 
Posts: 4764
Joined: 02 November 2005
Location: Southeastern USA

Re: Sudoku Explainer: Bugs, Quirks and Other Remarks...

Postby lksudoku » Fri Oct 08, 2010 11:52 pm

Looking at the SE source code, it seems that overlapping rate range for different techniques can cause a wrong rating for some puzzles

That is due to the way in which rating is calculated, suppose you have technique A with rating range from 6-7 and technique B with rating range 6.1-7.1, when the code for rating is performed, it may try technique A first, find a rating of 6.9 move and continue with that move, while it is possible that if technique B was tried first, it would yield a 6.2 rating for instance

To fix that, all overlapping rate techniques should be tried before choosing the next move
Last edited by lksudoku on Fri Oct 08, 2010 11:59 pm, edited 1 time in total.
lksudoku
 
Posts: 90
Joined: 06 October 2010

Re: Sudoku Explainer: Bugs, Quirks and Other Remarks...

Postby ronk » Fri Oct 08, 2010 11:58 pm

lksudoku wrote:all overlapping rate techniques should be tried before choosing the next move

As of now, we have no reason to believe it does otherwise.
ronk
2012 Supporter
 
Posts: 4764
Joined: 02 November 2005
Location: Southeastern USA

Re: Sudoku Explainer: Bugs, Quirks and Other Remarks...

Postby lksudoku » Sat Oct 09, 2010 12:00 am

ronk wrote:
lksudoku wrote:all overlapping rate techniques should be tried before choosing the next move

As of now, we have no reason to believe it does otherwise.

Yes you do, just look at the source code
lksudoku
 
Posts: 90
Joined: 06 October 2010

Re: Sudoku Explainer: Bugs, Quirks and Other Remarks...

Postby ronk » Sat Oct 09, 2010 12:08 am

lksudoku wrote:
ronk wrote:
lksudoku wrote:all overlapping rate techniques should be tried before choosing the next move

As of now, we have no reason to believe it does otherwise.

Yes you do, just look at the source code

LOL, I don't even believe my read of my own code ... let alone someone else's code. I'll believe your conjecture when we have a test example.
ronk
2012 Supporter
 
Posts: 4764
Joined: 02 November 2005
Location: Southeastern USA

Re: Sudoku Explainer: Bugs, Quirks and Other Remarks...

Postby lksudoku » Sat Oct 09, 2010 12:17 am

ronk wrote:LOL, I don't even believe my read of my own code ... let alone someone else's code. I'll believe your conjecture when we have a test example.


It can be tested easily, what you need is a puzzle which, for instance, can be solved with Forcing X-Chains relatively simply (that is in the lower end of the range 6.6 - 7.6), but when applying X-Cycles, Y-Cycles, it gets the higher end of the rating 6.5 - 7.5, How to generate such a puzzle, that is up to you, the same approach can be used for any other overlapping range

You may try to rate while disabling some earlier overlapping techniques, to find such a puzzle

I am quite certain of my reading of the code and do not intend to find such an example
lksudoku
 
Posts: 90
Joined: 06 October 2010

Re: Sudoku Explainer: Bugs, Quirks and Other Remarks...

Postby daj95376 » Sat Oct 09, 2010 5:58 am

I finally downloaded Sudoku Explainer, scanned the web site, and retrieved this ratings list.

    * 1.0: Last value in block, row or column
    * 1.2: Hidden Single in block
    * 1.5: Hidden Single in row or column
    * 1.7: Direct Pointing
    * 1.9: Direct Claiming
    * 2.0: Direct Hidden Pair
    * 2.3: Naked Single
    * 2.5: Direct Hidden Triplet
    * 2.6: Pointing
    * 2.8: Claiming
    * 3.0, 3.2, 3.4: Naked Pair, X-Wing, Hidden Pair
    * 3.6, 3.8, 4.0: Naked Triplet, Swordfish, Hidden Triplet
    * 4.2, 4.4: XY-Wing, XYZ-Wing
    * 4.5 - 5.0: Unique rectangles and loops
    * 5.0, 5.2, 5.4: Naked Quad, Jellyfish, Hidden Quad
    * 5.6 - 6.0: Bivalue Universal Graves
    * 6.2: Aligned Pair Exclusion
    * 6.5 - 7.5: Bidirectioal X-Cycles and Y-Cycles
    * 6.6 - 7.6: Forcing X-Chains
    * 7.0 - 8.0: Forcing Chains, Bidirectional Cycles
    * 7.5 - 8.5: Nishio
    * 8.0 - 9.0: Cell/Region Forcing Chains
    * 8.5 - 9.5: Dynamic Forcing Chains
    * 9.0 - 10.0: Dynamic Forcing Chains (+)
    * > 9.5: Nested Forcing Chains
I then had SE rate ronk's first two puzzles. The only apparent difference was in the single highest level technique used. The issue of overlapping ratings is mute in the case of these two puzzles.

original:

Code: Select all
 1 . . . . . . . .
 . . 2 1 . . 3 . .
 . 3 . . 4 . . 2 .
 . 4 . 3 . 5 . . .
 . . 5 . 2 . 6 . .
 . . . 6 . . . 7 .
 . 2 . . 7 . . 8 .
 . . 3 . . 8 9 . .
 . . . . . . . . 4    # ED=4.6/2.5/2.5

This Sudoku can be solved using the following logical methods:

 57 x Hidden Single
  1 x Direct Hidden Pair
  1 x Direct Hidden Triplet
  2 x Pointing
  1 x Naked Pair
  1 x Unique Loop type 2

 Unique Loop indicated by Sudoku Explainer
 +-----------------------------------------------------------------------+
 |  1      56789  6789   |  25789  3      2679   |  4      569    567    |
 |  4      56789  2      |  1      5689   679    |  3      569    567    |
 |  5679   3      679    |  579    4      679    |  8      2      1      |
 |-----------------------+-----------------------+-----------------------|
 |  67     4      67     |  3     *89     5      |  2      1     *89     |
 |  3     *89+1   5      |  789    2      179    |  6      4     *89     |
 |  2     *89+1   89-1   |  6     *89+1   4      |  5      7      3      |
 |-----------------------+-----------------------+-----------------------|
 |  569    2      4      |  59     7      3      |  1      8      56     |
 |  567    1567   3      |  4      156    8      |  9      56     2      |
 |  8      1569   169    |  259    1569   1269   |  7      3      4      |
 +-----------------------------------------------------------------------+
 # 82 eliminations remain

morph:

Code: Select all
 . 8 . . 9 . 3 . .
 . . . 4 . . . . .
 . . 7 . . 8 . 2 .
 . . 4 . . 2 . 3 .
 . . . . . . . . 1
 1 . . . 3 . 2 . .
 3 5 . . . . . 4 .
 . . 2 . 6 . 5 . .
 6 . . . . 7 . . .    # ED=6.6/2.5/2.5

This Sudoku can be solved using the following logical methods:

 57 x Hidden Single
  1 x Direct Hidden Pair
  1 x Direct Hidden Triplet
  2 x Pointing
  1 x Naked Pair
  1 x Turbot Fish

 +-----------------------------------------------------------------------+
 |  4      8      156    |  2      9      56     |  3      1567   567    |
 |  259    1269   1569   |  4      7      3      |  169    1569   8      |
 |  59     3      7      |  56     1      8      |  4      2      569    |
 |-----------------------+-----------------------+-----------------------|
 |  579    679    4      |  1      8      2      |  679    3      5679   |
 |  25789  2679   3      |  567    4      569    |  6789   56789  1      |
 |  1      679    5689   |  567    3      569    |  2      56789  4      |
 |-----------------------+-----------------------+-----------------------|
 |  3      5     *89     | *89     2      1      |  67     4      67     |
 |  789    179    2      | *89     6      4      |  5     *89+1   3      |
 |  6      4    *89+1    |  3      5      7      |  89-1  *89+1   2      |
 +-----------------------------------------------------------------------+
 # 82 eliminations remain

However, a unique loop still exists (*) in the second grid. It may or may not correspond to a "type 2".

The only (apparent) way to get the original puzzle and it's morph to have the same rating would be to convert them to MinLex form and rate it. Then that rating could be assigned to the original and morph puzzles.

Code: Select all
 Puzzle based on row MinLex of solution grid
 +-----------------------+
 | 1 . . | . 5 . | 7 . . |
 | . 5 . | 1 . . | . . 6 |
 | . . . | . . 7 | . . . |
 |-------+-------+-------|
 | 2 . . | . 1 . | . 4 . |
 | . . 6 | . . . | . . . |
 | . 4 . | 5 . . | . . 3 |
 |-------+-------+-------|
 | . . . | 6 . . | 1 9 . |
 | 8 . . | . 9 . | . . 5 |
 | . 3 . | . . . | 8 . . |
 +-----------------------+

This Sudoku can be solved using the following logical methods:

 57 x Hidden Single
  1 x Direct Hidden Pair
  1 x Direct Hidden Triplet
  2 x Pointing
  1 x Naked Pair
  1 x Unique Loop type 2

Regards, Danny
daj95376
2014 Supporter
 
Posts: 2624
Joined: 15 May 2006

Re: Sudoku Explainer: Bugs, Quirks and Other Remarks...

Postby lksudoku » Sat Oct 09, 2010 9:54 am

It is not possible to see the overlapping problem by using only SE as it is for rating, since it will always rate in the same way (same order of techniques), to view an example of the problem you need a "correct" rater and compare it to the current SE, such a "correct" rater can be achieved by one of the followings:
- Altering the SE
- Using an external rater
- Manually solve and rate
lksudoku
 
Posts: 90
Joined: 06 October 2010

Re: Sudoku Explainer: Bugs, Quirks and Other Remarks...

Postby lksudoku » Sat Oct 09, 2010 10:55 am

daj95376 wrote:However, a unique loop still exists (*) in the second grid. It may or may not correspond to a "type 2".

This may imply that there is an additional bug which causes the solver not to find unique loops that exist, however this bug I did not verify in the source code
lksudoku
 
Posts: 90
Joined: 06 October 2010

Re: Sudoku Explainer: Bugs, Quirks and Other Remarks...

Postby ronk » Sat Oct 09, 2010 6:45 pm

When the rating for a puzzle and its transpose differ, there must be a design error or a coding error.

Code: Select all
original:
 1 . . . . . . . .
 . . 2 1 . . 3 . .
 . 3 . . 4 . . 2 .
 . 4 . 3 . 5 . . .
 . . 5 . 2 . 6 . .
 . . . 6 . . . 7 .
 . 2 . . 7 . . 8 .
 . . 3 . . 8 9 . .
 . . . . . . . . 4    ED=4.6/2.5/2.5

transpose:
 1 . . . . . . . .
 . . 3 4 . . 2 . .
 . 2 . . 5 . . 3 .
 . 1 . 3 . 6 . . .
 . . 4 . 2 . 7 . .
 . . . 5 . . . 8 .
 . 3 . . 6 . . 9 .
 . . 2 . . 7 8 . .
 . . . . . . . . 4    ED=6.6/2.5/2.5
ronk
2012 Supporter
 
Posts: 4764
Joined: 02 November 2005
Location: Southeastern USA

Re: Sudoku Explainer: Bugs, Quirks and Other Remarks...

Postby 999_Springs » Sat Oct 09, 2010 8:33 pm

lksudoku wrote:all overlapping rate techniques should be tried before choosing the next move


ronk wrote:I'll believe your conjecture when we have a test example.

If you read my original post, this is exactly what I mentioned in #67. And it has nothing to do with the glitch in #40.
Hidden Text: Show
Once upon a time I was a teenager who was active on here 2007-2011
999_Springs
 
Posts: 367
Joined: 27 January 2007
Location: In the toilet, flushing down springs, one by one.

Next

Return to Software