SudokuSolver Forum

A forum for Sudoku enthusiasts to share puzzles, techniques and software
It is currently Sat Apr 27, 2024 9:52 pm

View unanswered posts | View active topics


Board index » Puzzles and Solutions » Killer Puzzles

All times are UTC




Post new topic Reply to topic  Page 1 of 1
  [ 7 posts ] 
  Print view Previous topic | Next topic 
Author Message
Børge
 Post subject: goooders' Special 1
PostPosted: Thu Oct 21, 2010 6:19 pm 
Offline
Grand Master
Grand Master
User avatar

Joined: Mon Apr 21, 2008 10:32 am
Posts: 868
Unless there are MASSIVE PROTEST, there will be an upcoming Assassin having puzzles with exactly this cage pattern.
I have 9 puzzles with a good SS score spread from 0.93 through 7.02, which both JSudoku and SudokuSolver can solve.
Have to do some analysis of the puzzles first to select which ones to post.

According to Ed one of the characteristics of a good Assassin is that SudokuSolver can solve it in a blink of an eye, and that JSudoku has a terrible time solving it.
The v1 at 1.40 of the upcoming Assassin has exactly this characteristic.

The following puzzle is the exact opposite. SudokuSolver rates it at 7.02, and JSudoku solves it instantly.
Based on the techniques JSudoku uses, my ballpark figure rating is Hard 1.75.
What does the human solvers say?

EDIT: "The 45 Rule lovers" will hopefully NOT get a raw deal: SudokuSolver uses fifteen 45 rules distributed over nine different 45 rule techniques.

normalized PS-code:
3x3::k:2832:3851:5132:5132:5132:5132:7939:7939:7939:2832:3851:3851:5132:8966:8966:9729:7939:7939:2575:3851:11522:8966:8966:9729:9729:9729:7939:2575:3851:11522:8966:9729:9729:9729:7173:4618:4113:4113:11522:11522:11522:9729:7173:7173:4618:9735:8964:8964:8964:11522:7173:7173:4618:4618:9735:9735:8964:8964:11522:11522:11522:4105:4618:3080:9735:9735:8964:2322:4105:4105:4105:4105:3080:3080:9735:9735:2322:3091:3091:4110:4110:
Solution: 
+-------+-------+-------+
| 8 2 6 | 1 4 7 | 3 9 5 |
| 3 4 1 | 2 5 9 | 8 6 7 |
| 9 5 7 | 6 8 3 | 2 4 1 |
+-------+-------+-------+
| 1 3 2 | 7 6 5 | 9 8 4 |
| 7 9 8 | 4 3 1 | 6 5 2 |
| 4 6 5 | 8 9 2 | 7 1 3 |
+-------+-------+-------+
| 2 7 4 | 9 1 6 | 5 3 8 |
| 5 8 9 | 3 7 4 | 1 2 6 |
| 6 1 3 | 5 2 8 | 4 7 9 |
+-------+-------+-------+

Image     Image     Image


Last edited by Børge on Fri Oct 22, 2010 4:48 pm, edited 1 time in total.

Top
 Profile  
Reply with quote  
Andrew
 Post subject: Re: goooders' Special 1
PostPosted: Fri Oct 22, 2010 12:15 am 
Offline
Grand Master
Grand Master

Joined: Wed Apr 23, 2008 6:04 pm
Posts: 1895
Location: Lethbridge, Alberta, Canada
Børge wrote:
According to Ed one of the characteristics of a good Assassin is that SudokuSolver can solve it in a blink of an eye, and that JSudoku has a terrible time solving it.
The v1 at 1.40 of the upcoming Assassin has exactly this characteristic.
I've never downloaded any solver so have no experience of using them. It seems strange to me that one solver can solve puzzles very quickly while the other struggles, even though I've read that one is better at 45s and combination work while the other is better for chains, "fish" and the like.

Børge wrote:
The following puzzle is the exact opposite. SudokuSolver rates it at 7.02, and JSudoku solves it instantly.
Based on the techniques JSudoku uses, my ballpark figure rating is Hard 1.75.
What does the human solvers say?
Don't know whether I'll try that one; it depends whether I find the time and inclination to try it. I welcome puzzles with this cage pattern including the planned v1 with a score of 1.40. Maybe that will be A203? I hope that someone else will claim A202 since Børge posted the A201 puzzles.
Top
 Profile  
Reply with quote  
Børge
 Post subject: Re: goooders' Special 1
PostPosted: Fri Oct 22, 2010 4:44 pm 
Offline
Grand Master
Grand Master
User avatar

Joined: Mon Apr 21, 2008 10:32 am
Posts: 868
Andrew wrote:
It seems strange to me that one solver can solve puzzles very quickly while the other struggles, even though I've read that one is better at 45s and combination work while the other is better for chains, "fish" and the like.
My understandig is as follows:
JSudoku has solvers for X techniques. SodokuSolver has solver for Y techniques.
if you take a ∆X = ∆Y and for both ∆X and ∆Y sort the techniques to be deployed in the same order, I would assume that both solvers will perform somewhat equal for most Killers.
The default sequence of solvers' deployment is however pretty different for the two solvers.

I am far from an expert when it comes to Killer techniques, but to me it looks like JSudoku uses a technique for solving this puzzles, which is not a subset of general chains, and for which SudokuSolver does not have a solver, so it has to resort to using a very large number of Bowmans Bingos:
List of Bowmans Bingos used by SudokuSolver:
2 Bowmans Bingo Complex 20
4 Bowmans Bingo Complex 18
2 Bowmans Bingo Complex 17
1 Bowmans Bingo Complex 16
1 Bowmans Bingo Complex 14
1 Bowmans Bingo Complex 13
1 Bowmans Bingo Complex 12
2 Bowmans Bingo Complex 8
2 Bowmans Bingo Complex 5
2 Bowmans Bingo Simple (Lite) 1
5 Bowmans Bingo Extended (Lite) 4
2 Bowmans Bingo Extended (Lite) 3
5 Bowmans Bingo Extended (Lite) 2

Andrew wrote:
Don't know whether I'll try that one; it depends whether I find the time and inclination to try it.
You absolutely should try it.
If you solve it, you can boast about having solved a Killer rated 7.02 by SudokuSolver ;)
Do not think that anyone has ever solved a Killer rated 7.02 by SudokuSolver.


I am also trying to see if JSudoku can generate a puzzle based on this cage pattern with large cages only, but I have now pretty much given this up:
Hidden Text:
Screenshot 1
Image



Screenshot 2
Image
Top
 Profile  
Reply with quote  
Andrew
 Post subject: Re: goooders' Special 1
PostPosted: Mon Nov 15, 2010 7:32 pm 
Offline
Grand Master
Grand Master

Joined: Wed Apr 23, 2008 6:04 pm
Posts: 1895
Location: Lethbridge, Alberta, Canada
Andrew wrote:
Don't know whether I'll try that one; it depends whether I find the time and inclination to try it.
Børge wrote:
You absolutely should try it.
If you solve it, you can boast about having solved a Killer rated 7.02 by SudokuSolver ;)
Do not think that anyone has ever solved a Killer rated 7.02 by SudokuSolver.
Many thanks Børge for persuading me to have a go at this one, after the hard slog of A199V2.

This was a fun puzzle. :D I guess I could call it a One (and a half) Trick Pony, step 9a being the main breakthrough and step 18 leading to a much quicker finish.

Børge wrote:
"The 45 Rule lovers" will hopefully NOT get a raw deal: SudokuSolver uses fifteen 45 rules distributed over nine different 45 rule techniques.
I didn't try finding them all. I only used two of them and spotted two others but in those cases immediately found quicker ways (steps 2 and 4) to get better results.

Rating Comment:
I'll rate this puzzle at 1.5. It was hard to decide a rating for step 9 but since it's basically a "sees all except" step it feels like 1.5. I also used a locking-out cages step and a couple of caged X-Wings. There's also a killer quint but it's so obvious that I can't give it a high rating.

Here is my walkthrough:
Prelims

a) R12C1 = {29/38/47/56}, no 1
b) R34C1 = {19/28/37/46}, no 5
c) R5C12 = {79}
d) R89C5 = {18/27/36/45}, no 9
e) R9C67 = {39/48/57}, no 1,2,6
f) R9C89 = {79}
g) 15(5) cage at R1C2 = {12345}
h) 35(5) cage at R2C5 = {56789}
i) 18(5) cage at R4C9 = {12348/12357/12456}, no 9
j) 16(5) cage at R7C8 = {12346}
k) 38(8) cage at R2C7 = {12345689}, no 7
l) And, of course, 45(9) cage at R3C3 = {123456789}

Steps resulting from Prelims
1a. Naked pair {79} in R5C12, locked for R5 and N4, clean-up: no 1,3 in R3C1
1b. Naked pair {79} in R9C89, locked for R9 and N9, clean-up: no 2 in R8C5, no 3,5 in R9C78
1c. Naked pair {48} in R9C67, locked for R9, clean-up: no 1,5 in R8C5
1d. 5 in N9 only in R7C79, locked for R7
1e. Naked quint {56789} in 35(5) cage at R2C5, CPE no 5,6,7,8,9 in R12C4

2. Killer quint 1,2,3,4,5 in R12C1 and 15(5) cage at R1C2, locked for N1, clean-up: no 6,8 in R4C1
2a. 1 in N1 only in 15(5) cage at R1C2 -> no 1 in R4C2

3. 45 rule on N2 2(1+1) outies R1C3 + R4C4 = 1 innie R3C6 + 10
3a. Max R1C3 + R4C4 = 18 -> max R3C6 = 8

4. 16(5) cage at R7C8 = {12346}, R9C67 = {48} -> caged X-Wing for 4, no other 4 in C6, N8 and N9, clean-up: no 5 in R9C5

5. 5 in N8 only inR89C4, locked for C4
5a. 35(5) cage at R2C5 = {56789}, 5 locked for N2
5b. Min R1C3 + R4C4 = 12 -> min R3C6 = 2 (step 3)
5c. 5 in R9 only in R9C1234, CPE no 5 in R8C23
5d. 6,8 in N4 only in R45C3 + R6C123, CPE no 6,8 in R6C5

6. 38(7) cage at R6C1 = {1256789/1346789/2345789}, 7,9 locked for N7
6a. 8 only in R6C1 + R7C12 + R8C23, CPE no 8 in R8C1

7. 12(3) cage in N7 = {156/246/345}
7a. 4 of {246/345} must be in R8C1 -> no 2,3 in R8C1

8. 35(6) cage at R6C2 = {146789/236789/245789/345689}, 9 locked for C4
8a. 35(5) cage at R2C5 = {56789}, 9 locked for N2

9. R1C3 + R4C4 = R3C6 + 10 -> R3C6 must be less than either of R1C3 and R4C4
9a. R1C3 and R4C4 cannot have the same value because they “see” all of N2 except for R3C6, which cannot equal either R1C3 or R4C4 -> R1C3 + R4C4 cannot be [66] = 12 -> no 2 in R3C6

10. 1,2,4 in N2 only in 20(5) cage at R1C3 = {12467}, 6,7 locked for R1, clean-up: no 4,5 in R2C1

11. R3C6 = 3 (hidden single in N2)
11a. R1C3 + R4C4 = R3C6 + 10 (step 3)
11b. R3C6 = 3 -> R1C3 + R4C4 = 13 = {67}, CPE no 6 in R4C4

12. 16(5) cage at R7C8 = {12346}, 3 locked for N9

13. 45 rule on N3 3 innies R2C7 + R3C78 = 14 = {149/158/248}, no 6

14. R6C7 = 7 (hidden single in C7)
14a. 18(5) cage at R4C9 = {12348/12456}, 4 locked for N6

15. R4C4 = 7 (hidden single in R4) -> R1C3 = 6 (step 11b), clean-up: no 5 in R1C1, no 4 in R4C1
15a. 5 in N1 only in 15(5) cage at R1C2 -> no 5 in R4C2

16. Killer triple 7,8,9 in R12C1, R3C1 and R5C1, locked for C1
16a. 38(7) cage at R6C1 = {1256789/1346789/2345789} -> R7C2 + R8C23 = {789}, 8 locked for N7
16b. Naked triple {789} in R578C2, 8 locked for C2 and N7

17. R38C3 = {79} (hidden pair in C3)

[I first saw the next step as a contradiction chain which showed that R4C2 cannot be 2, next I found a shorter contradiction chain
R12C1 = {29/38/47} cannot be {29}, here’s how
R12C1 = {29} => R34C1 = [73] clashes with R3C3 = 7
-> R12C1 = {38/47}, no 2,9
Later I found a more satisfying way to do it]

18. 8 in C1 only in R12C1 = {38} or R34C1 = [82] -> R12C1 = {38/47} (cannot be {29}, locking-out cages), no 2,9
18a. 2 in N1 only in 15(5) cage at R1C2 -> no 2 in R4C2
18b. 9 in N1 only in R3C13, locked for R3
18c. 9 in N2 only in R2C56, locked for R2

19. R2C7 + R3C78 (step 13) = {158/248}, 8 locked for N3 and 38(7) cage at R2C7, no 8 in R4C567 + R5C6

20. R1C1 = 8 (hidden single in R1), R2C1 = 3, clean-up: no 7 in R3C1, no 2 in R4C1
20a. 15(5) cage at R1C2 = {12345} -> R4C2 = 3

21. R34C1 = [91], R5C12 = [79], R3C3 = 7, R8C3 = 9

22. 12(3) cage in N7 (step 7) = {156} (cannot be {246} which clashes with R7C1) -> R9C2 = 1, R89C1 = {56}, locked for C1 and N7, clean-up: no 8 in R8C5

23. 38(7) cage at R6C1 (step 16a) = {2345789} (only remaining combination) -> R9C3 = 3, R9C4 = 5, R89C1 = [56], R9C5 = 2, R8C5 = 7, R78C2 = [78]
23a. Naked pair {24} in R7C13, locked for R7
23b. R1C6 = 7 (hidden single in R1)
23c. 2 in N2 only in R12C4, locked for C4

24. Naked triple {245} in R123C2, locked for C2 and N1 -> R2C3 = 1, R6C2 = 6
24a. 1 in N2 only in R1C45, locked for R1

25. 38(8) cage at R2C7 = {12345689}, 9 locked for R4
25a. R4C7 = 9 (hidden single in N6
25b. 38(8) cage at R2C7 = {12345689}, 6 locked for N5

26. 2 in 45(9) cage at R3C3 only in R45C3, locked for C3 and N4 -> R67C1 = [42], R7C3 = 4

27. 35(6) cage at R6C2 (step 8) = {345689} (only remaining combination) -> R8C4 = 3, R6C3 = 5, R67C4 = {89}, locked for C4 -> R3C4 = 6
27a. R7C8 = 3 (hidden single in R7)

28. Naked pair {28} in R45C3, locked for 45(9) cage at R3C3, no 8 in R5C5 + R7C567
28a. 5 in 45(9) cage only in R5C5 + R7C7, CPE no 5 in R5C7

29. R67C4 = {89}, 9 in 45(9) cage at R3C3 only in R6C5 + R7C56 -> caged X-Wing for 9 in R67, no other 9 in R6
[Alternatively this elimination of 9 from R6C6 can be made using 9 in R2 only in R2C56 with a caged X-Wing in C56.]

30. 28(5) cage at R4C8 = {25678} (only remaining combination containing 7 but neither of 4,9), 5,6 locked for N6, 5 also locked for C8, CPE no 2,8 in R6C89 -> R6C89 = [13], R6C5 = 9, R67C4 = [89], R6C6 = 2

31. Naked pair {16} in R7C56, locked for R7, N8 and 45(9) cage at R3C3 -> R7C7 = 5

and the rest is naked singles.

If it was programmed in, SudokuSolver could potentially have made my breakthrough step this way:
After reducing R1C3 and R4C4 to {6789}
45 rule on N2 2(1+1) outies R1C3 + R4C4 = 1 innie R3C6 + 10
R1C3 + R4C4 cannot be [66] = 12 => R3C6 = 2 and cannot place 6 in N2


Last edited by Andrew on Mon Nov 15, 2010 7:48 pm, edited 2 times in total.

Top
 Profile  
Reply with quote  
Ed
 Post subject: Re: goooders' Special 1
PostPosted: Tue Nov 16, 2010 9:18 am 
Offline
Grand Master
Grand Master

Joined: Wed Apr 16, 2008 1:16 am
Posts: 1044
Location: Sydney, Australia
Børge wrote:
there will be an upcoming Assassin having puzzles with exactly this cage pattern.
Love the look of that cage pattern. Look forward to it.

Well done Andrew for taking on this goooders special! The key step SS can't get is your step 4. I've sent it to Richard hoping it's just a bug that can be easily fixed. I know it can do that type of step (grouped X-cycle) at other times so fingers crossed.

Cheers
Ed
Top
 Profile  
Reply with quote  
Børge
 Post subject: Re: goooders' Special 1
PostPosted: Thu Jun 30, 2011 10:24 pm 
Offline
Grand Master
Grand Master
User avatar

Joined: Mon Apr 21, 2008 10:32 am
Posts: 868
Andrew, Thanks a lot for giving this puzzle a try :sun:
You deserve a place in Guinness Book of World Records for having solved a Killer rated 7.02 by SudokuSolver. :applause: :cheers:

Børge wrote:
... there will be an upcoming Assassin having puzzles with exactly this cage pattern.
If there is still interest, I will make good on my long due promise.
Top
 Profile  
Reply with quote  
Andrew
 Post subject: Re: goooders' Special 1
PostPosted: Sat Jul 02, 2011 4:27 am 
Offline
Grand Master
Grand Master

Joined: Wed Apr 23, 2008 6:04 pm
Posts: 1895
Location: Lethbridge, Alberta, Canada
Thanks Børge for your nice comments! :)

I look forward to your offer of posting an Assassin with this cage pattern. From his post in this thread, I'm sure Ed will look forward to it also. Ed loves puzzles with large cages.

A puzzle with this cage pattern and SSscore of 1.40 should make a good Assassin.
Top
 Profile  
Reply with quote  
Display posts from previous:  Sort by  
Post new topic Reply to topic  Page 1 of 1
  [ 7 posts ] 

Board index » Puzzles and Solutions » Killer Puzzles

All times are UTC


Who is online

Users browsing this forum: No registered users and 47 guests

You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot post attachments in this forum

Search for:
Jump to:  
Powered by phpBB® Forum Software © phpBB Group