The hardest sudokus (new thread)

Everything about Sudoku that doesn't fit in one of the other sections

Re: The hardest sudokus (new thread)

Postby hendrik_monard » Mon Apr 19, 2021 8:41 pm

Hi, this is my first input to this forum. I've heen following it for some time and I like the cooperative atmosphere.
I think I have a new 11.8
I made a canonical form of it and could not find it in the index of the ph_2010 database:
98.7.....6...5.8....4....3.7..9..5....3....4......2..1.6...1..2..9.........89.6.. 11.8/1.2/1.2
Can anyone confirm this rating?
hendrik_monard
 
Posts: 96
Joined: 19 April 2021
Location: Leuven (Louvain) Belgium

Re: The hardest sudokus (new thread)

Postby mith » Fri Apr 23, 2021 7:06 pm

Looks like it's new and minimal, and I got 11.8 for it as well. Nice find! :)
mith
 
Posts: 996
Joined: 14 July 2020

Re: The hardest sudokus (new thread)

Postby denis_berthier » Sat Apr 24, 2021 3:28 am

hendrik_monard wrote:Hi, this is my first input to this forum. I've heen following it for some time and I like the cooperative atmosphere.
I think I have a new 11.8
I made a canonical form of it and could not find it in the index of the ph_2010 database:
98.7.....6...5.8....4....3.7..9..5....3....4......2..1.6...1..2..9.........89.6.. 11.8/1.2/1.2
Can anyone confirm this rating?

Good find.
As usual when a very high rank appears (which is quite rare now), I checked its BpB classification: it is (exactly) in B5B.
The club of B7B remains restricted to its old 3 members.
denis_berthier
2010 Supporter
 
Posts: 4213
Joined: 19 June 2007
Location: Paris

PGExplainer.jar

Postby 1to9only » Thu May 27, 2021 2:13 pm

::: shameless plug :::

You should really be using my PGExplainer - a Minimal SudokuExplainer, in 56,712 bytes to rate (ER/EP/ED) these hard puzzles.

Time taken to rate 999_Spring's 42 clues ED=11.1/11.1/2.8 puzzle posted Mar 24, 2021 here:
Code: Select all
                        Version     Size            Rating Time

SudokuExplainer.jar   - 1.2.1.3   - 304,777 bytes -      59 s

PGExplainer.jar       - 2021.3.30 - 56,712 bytes  -      16 s


Time taken to rate mith's 9 26 clues ED=11.6/1.2/1.2 puzzles posted Apr 06, 2021 here:
Code: Select all
                        Version     Size            Rating Time

PGExplainer.jar       - 2021.3.30 - 56,712 bytes  - 21 m 20 s   - average: 2 m 22 s per puzzle

User avatar
1to9only
 
Posts: 4177
Joined: 04 April 2018

Re: The hardest sudokus (new thread)

Postby mith » Fri May 28, 2021 6:58 pm

New high for 28c minimal:

Code: Select all
........1.....2..3..4....5...5..4..6.6..785.27......8..5672....24..856..8.74.6...  ED=11.5/1.2/1.2 (DCFC+DFC)
........1.....2..3..4....5...5..4..6.7.....8.64..785.2.2..856...874.6...5.672....  ED=11.5/1.2/1.2 (DCFC+DFC)
Last edited by mith on Tue Jun 01, 2021 3:40 pm, edited 1 time in total.
mith
 
Posts: 996
Joined: 14 July 2020

Re: The hardest sudokus (new thread)

Postby mith » Tue Jun 01, 2021 3:39 pm

And 29c minimal:

Code: Select all
................12.....34.5..6.71..4.4783.....8.6.43.7.71..6...6.4.87...83.41.... ED=11.2/1.2/1.2 (DCFC+MFC)
mith
 
Posts: 996
Joined: 14 July 2020

Re: The hardest sudokus (new thread)

Postby mith » Sun Jun 06, 2021 4:09 am

New 11.8:

Code: Select all
........1....23.....45...2...3..5.6..6....1..78.6....9..6.5.....1.4....89.....7..  ED=11.8/11.8/2.6 (DCFC+DFC)
mith
 
Posts: 996
Joined: 14 July 2020

Re: The hardest sudokus (new thread)

Postby denis_berthier » Sun Jun 06, 2021 10:03 am

mith wrote:New 11.8:

Code: Select all
........1....23.....45...2...3..5.6..6....1..78.6....9..6.5.....1.4....89.....7..  ED=11.8/11.8/2.6 (DCFC+DFC)

Congrats for a new 11.8.
I checked its BpB classification: B6B (exactly).
denis_berthier
2010 Supporter
 
Posts: 4213
Joined: 19 June 2007
Location: Paris

Re: The hardest sudokus (new thread)

Postby 999_Springs » Fri Jun 11, 2021 11:04 am

hey 1to9only

from your thread in the interactive games section here, you've found a way to get SE to report the "node" count of a particular forcing chain, which is what SE uses to calculate a difficulty rating for that chain

it's my understanding that when SE takes a step, it'll find all chains within a certain category and choose the shortest one among them, even compared to other chains with the same rating that may have slightly higher node counts, and it'll use that as its next step (someone confirm or deny this)

because of this, we can get a more detailed ranking of hardest puzzles in the 11.7+ range (nested fc + dynamic fc) by taking the highest node count of the hardest step in each puzzle. for instance, in your latest post you said that Mike's 20 clue 11.8 has a 318 node nested dynamic chain

would it be possible to get a complete ranking of the 11.7+'s by node count of the longest nested dynamic chain? i think this would be a more satisfying tie-breaker for the hardest puzzles by se rating, rather than the traditionally used ep and ed ratings. it would be interesting to see at least which is the hardest se 11.9 :)

this is only well defined for 11.7+ because for 11.6 you start to run into the next category below (nested multiple fc) and you have no way to compare the difficulty of a longer nested multiple fc compared to a shorter nested dynamic fc

thanks
999_Springs
 
Posts: 591
Joined: 27 January 2007
Location: In the toilet, flushing down springs, one by one.

Re: The hardest sudokus (new thread)

Postby 1to9only » Fri Jun 11, 2021 12:43 pm

Example used (highest rated in current PG422):
Code: Select all
100008007006100400050030080060300004009000600400006070005020010200001700010900002 #  50 10.7/10.7/10.6

999_Springs wrote:from your thread in the interactive games section here, you've found a way to get SE to report the "node" count of a particular forcing chain, which is what SE uses to calculate a difficulty rating for that chain

My version of SukakuExplainer: here has been modified to get (and display) the chains nodes count (used to calculate difficulty) - the info appears in the Hints description on the right of the SE window. 1st step:
Code: Select all
Contradiction Forcing Chain (w/148 nodes): R5C1.7 on ==> R7C9.6 both on & off

The 'Analyse (F9)' does not display the chains nodes info, as it groups the techniques used together.

999_Springs wrote:it's my understanding that when SE takes a step, it'll find all chains within a certain category and choose the shortest one among them, even compared to other chains with the same rating that may have slightly higher node counts, and it'll use that as its next step (someone confirm or deny this)

Yes, this is so. 'Get all hints' will show the higher nodes hints. For the possible 1st step: 10.6-10.8 hints:
Hidden Text: Show
Code: Select all
Contradiction Forcing Chain (w/148 nodes): R5C1.7 on ==> R7C9.6 both on & off
Contradiction Forcing Chain (w/159 nodes): R7C6.7 on ==> R4C8.5 both on & off
Contradiction Forcing Chain (w/310 nodes): R7C1.7 on ==> R1C5.9 both on & off
Contradiction Forcing Chain (w/340 nodes): R2C9.9 on ==> R6C4.8 both on & off

Also 10.8-11.5 hints:
Hidden Text: Show
Code: Select all
Cell Forcing Chains (w/63 nodes): R2C9 ==> R1C8.5 off
Contradiction Forcing Chain (w/120 nodes): R1C8.3 on ==> R4C8.9 both on & off
Contradiction Forcing Chain (w/123 nodes): R4C8.5 on ==> R4C5.9 both on & off
Cell Forcing Chains (w/127 nodes): R5C8 ==> R1C8.3 off
Contradiction Forcing Chain (w/138 nodes): R1C8.5 on ==> R9C1.3 both on & off
Contradiction Forcing Chain (w/141 nodes): R1C8.9 on ==> R9C1.3 both on & off
Contradiction Forcing Chain (w/148 nodes): R5C1.7 on ==> R7C9.6 both on & off
Contradiction Forcing Chain (w/153 nodes): R9C1.3 on ==> R6C4.5 both on & off
Cell Forcing Chains (w/155 nodes): R2C9 ==> R5C9.5 off
Cell Forcing Chains (w/156 nodes): R5C8 ==> R3C4.7 off
Contradiction Forcing Chain (w/159 nodes): R7C6.7 on ==> R4C8.5 both on & off
Contradiction Forcing Chain (w/164 nodes): R3C4.7 on ==> R9C6.3 both on & off
Cell Forcing Chains (w/174 nodes): R5C8 ==> R5C5.7 off
Contradiction Forcing Chain (w/188 nodes): R3C9.9 on ==> R2C1.3 both on & off
Cell Forcing Chains (w/190 nodes): R5C8 ==> R8C2.8 off
Cell Forcing Chains (w/213 nodes): R5C8 ==> R7C1.7 off
Contradiction Forcing Chain (w/223 nodes): R2C5.9 on ==> R9C1.3 both on & off
Contradiction Forcing Chain (w/230 nodes): R3C7.9 on ==> R4C5.8 both on & off
Cell Forcing Chains (w/232 nodes): R5C8 ==> R7C2.8 off
Contradiction Forcing Chain (w/244 nodes): R5C6.7 on ==> R5C1.3 both on & off
Contradiction Forcing Chain (w/252 nodes): R5C5.7 on ==> R4C3.2 both on & off
Contradiction Forcing Chain (w/259 nodes): R5C9.5 on ==> R6C2.3 both on & off
Contradiction Forcing Chain (w/268 nodes): R8C8.3 on ==> R8C4.5 both on & off
Contradiction Forcing Chain (w/287 nodes): R7C2.8 on ==> R9C8.4 both on & off
Contradiction Forcing Chain (w/310 nodes): R7C1.7 on ==> R1C5.9 both on & off
Contradiction Forcing Chain (w/340 nodes): R2C9.9 on ==> R6C4.8 both on & off
Contradiction Forcing Chain (w/395 nodes): R2C2.9 on ==> R4C3.8 both on & off
Contradiction Forcing Chain (w/411 nodes): R9C1.8 on ==> R4C6.9 both on & off
Contradiction Forcing Chain (w/413 nodes): R7C1.3 on ==> R6C9.5 both on & off
Cell Forcing Chains (w/457 nodes): R2C9 ==> R6C9.5 off
Cell Forcing Chains (w/468 nodes): R2C9 ==> R2C8.5 off
Contradiction Forcing Chain (w/551 nodes): R8C8.5 on ==> R6C3.8 both on & off
Contradiction Forcing Chain (w/601 nodes): R8C2.8 on ==> R2C5.5 both on & off

999_Springs wrote:would it be possible to get a complete ranking of the 11.7+'s by node count of the longest nested dynamic chain? i think this would be a more satisfying tie-breaker for the hardest puzzles by se rating, rather than the traditionally used ep and ed ratings. it would be interesting to see at least which is the hardest se 11.9 :)

The chains nodes info is produced by obtaining the solution paths of the Patterns Game submissions, and collating them.
Code: Select all
java.exe -Xrs -Xmx1g -cp SukakuExplainer.jar diuf.sudoku.test.hints --input=puzzles.txt

The ED>10.0 (1st 6) steps for the ED=10.7 example are below:
Code: Select all
10.6, Contradiction Forcing Chain (w/148 nodes): R5C1.7 on ==> R7C9.6 both on & off: r5c1<>7
10.6, Contradiction Forcing Chain (w/159 nodes): R7C6.7 on ==> R4C8.5 both on & off: r7c6<>7
10.6, Contradiction Forcing Chain (w/180 nodes): R7C2.8 on ==> R5C8.5 both on & off: r7c2<>8
10.7, Contradiction Forcing Chain (w/193 nodes): R7C1.3 on ==> R6C9.5 both on & off: r7c1<>3
10.3, Cell Forcing Chains (w/57 nodes): R2C9 ==> R1C8.5 off: r1c8<>5
10.5, Contradiction Forcing Chain (w/111 nodes): R9C1.3 on ==> R1C7.9 both on & off: r9c1<>3

I have the solution paths of the 11.7+ submissions (up to PG420) - I'll collect and post the info later.
For the 11.9s, I might do them some time later - or someone else can follow the process outlined above.
User avatar
1to9only
 
Posts: 4177
Joined: 04 April 2018

Patterns Game ED=11.7+

Postby 1to9only » Fri Jun 11, 2021 3:19 pm

I can only find x2 11.7s, and x3 11.8s submitted in the Patterns Game - they are all shown below.
[g.r.emlin flags the ED=11.8/11.8/3.4 as the 2nd posted with that rating! I cannot find the 1st one!!]

Ordered by the number of 11.8 and 11.7 solving steps (difficulty!) in the solution path:

Code: Select all
..1...2...9...8.5.6.......7.....3.4.....2.....8.9.5...7.......6.5.4...3...2...1.. # 0169   3  11.8/11.8/3.4

11.7, Region Forcing Chains (w/243 nodes): 6 in column ==> R5C2.4 off: r5c2<>4
11.8, Contradiction Forcing Chain (w/318 nodes): R2C5.4 on ==> R9C1.9 both on & off: r2c5<>4
11.8, Contradiction Forcing Chain (w/321 nodes): R5C3.7 on ==> R2C5.7 both on & off: r5c3<>7
11.8, Region Forcing Chains (w/336 nodes): 8 in block ==> R8C5.9 off: r8c5<>9
11.8, Contradiction Forcing Chain (w/337 nodes): R1C5.7 on ==> R5C2.7 both on & off: r1c5<>7
11.8, Cell Forcing Chains (w/368 nodes): R9C2 ==> R5C2.3 off: r5c2<>3

Code: Select all
.1.7....3....9..2...2...6...7...8.....6....4.5..1....7....4.9..3....7..5.8.3..... # 0344   2   11.8/1.2/1.2

11.7, Contradiction Forcing Chain (w/229 nodes): R6C6.3 on ==> R2C9.4 both on & off: r6c6<>3
11.8, Contradiction Forcing Chain (w/296 nodes): R8C4.8 on ==> R9C5.5 both on & off: r8c4<>8
11.8, Contradiction Forcing Chain (w/311 nodes): R3C8.9 on ==> R5C1.1 both on & off: r3c8<>9
11.2, Contradiction Forcing Chain (w/208 nodes): R9C6.1 on ==> R8C5.8 both on & off: r9c6<>1
11.8, Contradiction Forcing Chain (w/310 nodes): R9C6.2 on ==> R6C2.2 both on & off: r9c6<>2

The highest submitted ED=11.8/11.8/10.7 has only ONE 11.8 solving step! And 280 nodes, which is less than in the above two!!
Code: Select all
..1.5..6.6.......2.4....3...3.7.......6.1...8.....9...5...2...1.7...49....8....2. # 0349   4 11.8/11.8/10.7

11.8, Cell Forcing Chains (w/280 nodes): R9C1 ==> R8C9.3 off: r8c9<>3

Code: Select all
8.7..2....6.4.....1....97...3.....5.........12.1...9....8..71.....5....4....6..73 # 0193  98   11.7/1.2/1.2

11.7, Contradiction Forcing Chain (w/210 nodes): R2C8.2 on ==> R6C9.6 both on & off: r2c8<>2
11.7, Contradiction Forcing Chain (w/206 nodes): R4C9.6 on ==> R5C6.5 both on & off: r4c9<>6
11.7, Region Forcing Chains (w/211 nodes): 3 in column ==> R2C9.5 off: r2c9<>5
11.7, Contradiction Forcing Chain (w/237 nodes): R5C8.6 on ==> R2C1.9 both on & off: r5c8<>6
11.7, Contradiction Forcing Chain (w/230 nodes): R4C5.4 on ==> R4C7.6 both on & off: r4c5<>4
11.7, Contradiction Forcing Chain (w/251 nodes): R9C6.1 on ==> R1C7.6 both on & off: r9c6<>1

Code: Select all
9.7.2.....1...5...3..6..9....37..2..2.......1.8.....4...62..3.......7..5....4..8. # 0399   3   11.7/1.2/1.2

11.7, Cell Forcing Chains (w/212 nodes): R3C8 ==> R2C5.8 off: r2c5<>8
11.7, Contradiction Forcing Chain (w/212 nodes): R3C8.7 on ==> R2C3.4 both on & off: r3c8<>7
User avatar
1to9only
 
Posts: 4177
Joined: 04 April 2018

Re: The hardest sudokus (new thread)

Postby mith » Fri Jun 11, 2021 8:51 pm

999_Springs wrote:this is only well defined for 11.7+ because for 11.6 you start to run into the next category below (nested multiple fc) and you have no way to compare the difficulty of a longer nested multiple fc compared to a shorter nested dynamic fc

thanks


This actually isn't quite true now; I have two examples of 11.7s which SE is able to solve with D*FC+MFC:

Code: Select all
........1..2..3..4.5..6.27.....8.52.5..9.6...8.......6..76..4...6.....1.2...7.8.. ED=11.7/1.2/1.2 (DCFC+MFC)
........1.....2.3...435.6.....2.7..6..6.......8.56.3...43.....9.58......6..48.5.. ED=11.7/1.2/1.2 (DCFC+MFC)


I'd be curious what the output for these is, 1to9. (Or I'll try them out myself when I have more time!) I wouldn't be surprised if there isn't a lower rated +DFC which solves these, but I don't think there is a way to check for that in SE as is (since it will only find the +MFC steps until it runs out).
mith
 
Posts: 996
Joined: 14 July 2020

Re: The hardest sudokus (new thread)

Postby 1to9only » Fri Jun 11, 2021 9:53 pm

mith wrote:
Code: Select all
........1..2..3..4.5..6.27.....8.52.5..9.6...8.......6..76..4...6.....1.2...7.8.. ED=11.7/1.2/1.2 (DCFC+MFC)
........1.....2.3...435.6.....2.7..6..6.......8.56.3...43.....9.58......6..48.5.. ED=11.7/1.2/1.2 (DCFC+MFC)


The DCFC+MFC have a base ED=10.0 with +1.7 for chains complexity (i.e. total chains length - 2) in the range 1025-1536 nodes.

Code: Select all
........1..2..3..4.5..6.27.....8.52.5..9.6...8.......6..76..4...6.....1.2...7.8.. ED=11.7/1.2/1.2 (DCFC+MFC)

11.7, Contradiction Forcing Chain (w/1079 nodes): R1C4.5 on ==> R4C2.3 both on & off: r1c4<>5


Code: Select all
........1.....2.3...435.6.....2.7..6..6.......8.56.3...43.....9.58......6..48.5.. ED=11.7/1.2/1.2 (DCFC+MFC)

11.7, Contradiction Forcing Chain (w/1183 nodes): R2C1.1 on ==> R9C8.7 both on & off: r2c1<>1

User avatar
1to9only
 
Posts: 4177
Joined: 04 April 2018

Re: The hardest sudokus (new thread)

Postby mith » Fri Jun 11, 2021 11:18 pm

Nice. Presumably that 1183 node chain is the largest in a known puzzle (though it's possible some higher rated puzzle was just assumed to be +DFC when it is also +MFC). These are the highest ratings I am aware of for each level:

dfc (9.7 -> 1.2)
dfc+ (10.1 -> 1.1)
dfc+fc (11.1 -> 1.6)
dfc+mfc (11.7 -> 1.7)
dfc+dfc (11.9 -> 1.4)

In other news, some more high count minimal records:

Code: Select all
29c
................12.....34.5..6.71..4.4783....18.6.43...71..6...6.4.87...83.41.... ED=11.3/1.2/1.2 (DCFC+MFC)
................12.....34.5.16..7...74..86...8.341.....648.....1.87.43..37..61..4 ED=11.3/1.2/1.2 (DCFC+MFC)
................12.....34.5..6.71....876.4...41.83.....71..6..46.418.3..83..47... ED=11.3/1.2/1.2 (DCFC+MFC)
..............1.23..1...4.5..6.742...4.81.....876.2....741.6...6.248..1.8...27... ED=11.3/1.2/1.2 (DCFC+MFC)
..............1.23..1...4.5..6.742...876.....24.81.....741.6...6.248..1.8...27... ED=11.3/1.2/1.2 (DCFC+MFC)

30c
..............1.23..1...4.5..6.74....876.2...24.81.....741.6..26.248..1.8...27... ED=11.3/1.2/1.2 (DCFC+MFC)

31c
..............1.23..2.34....25...67..3..7.2.88.76....5.53..68..28..5.36.7.6....52 ED=11.1/1.2/1.2 (DCFC+MFC)
........1....12.3...3..4....15.6.7.3.7...561.3.6....58.58...36.6.15...8773..8.... ED=11.1/1.2/1.2 (DCFC+MFC)
..............1.23124.....5..6.7...8.784.2...41..68....42.1.7..6.17...8278..26... ED=11.1/1.2/1.2 (DLFC+MFC)
mith
 
Posts: 996
Joined: 14 July 2020

Re: The hardest sudokus (new thread)

Postby mith » Sat Jun 12, 2021 1:18 am

Well, after posting that I updated the database, and then when I restarted the SE script this was the first to be rated:

Code: Select all
29c
........1.....2.3....45.2.6..6.....7.37.8....12.6..3.8.72..81..3.1...8.268.....73  ED=11.5/1.2/1.2 (DCFC+DFC)
mith
 
Posts: 996
Joined: 14 July 2020

PreviousNext

Return to General