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 champagne » Sun Nov 28, 2010 9:25 pm

ronk wrote:
champagne wrote: checked your puzzle and I made a run including the former 11.8

009000400070300020800060007100800006000010070000056000300005001040000090002000700 11.8
003006080000100206600070005004000000300004090098020000030008060800000107000500000 11.5
...
If we now look for special tools

I think I already noticed that your 11.8 puzzle has something close to an EXOCET (r56c4). The second one has an EXOCET (r46c1).
None have the SK loop nor a rank0 logic (i did not check for more than 4 floors).

Using the adequate special tool, the print is significantly reduced, but still around 150K.

For the ER=11.5 puzzle, what "special tool" was used ... and what were the exclusions?


The special tool was just indicated above, it is an EXOCET based in r46c1. I just got it in my stat file, but I will dig in the print to morrow and give a more precise answer.

champagne
champagne
2017 Supporter
 
Posts: 7453
Joined: 02 August 2007
Location: France Brittany

Re: The hardest sudokus (new thread)

Postby eleven » Mon Nov 29, 2010 10:11 am

Thanks for the analysis, Champagne.
eleven
 
Posts: 3142
Joined: 10 February 2008

Re: The hardest sudokus (new thread)

Postby champagne » Mon Nov 29, 2010 10:59 am

champagne wrote:The special tool was just indicated above, it is an EXOCET based in r46c1. I just got it in my stat file, but I will dig in the print to morrow and give a more precise answer.

champagne


Code: Select all
||124579 12457 3     |249   459    6     |479   8     149   
||4579   4578  579   |1     34589  359   |2     347   6     
||6      1248  129   |23489 7      239   |349   134   5     
---------------------------------------------------------
||1257   12567 4     |36789 135689 13579 |35678 12357 1238 
||3      12567 12567 |678   1568   4     |5678  9     128   
||157    9     8     |367   2      1357  |34567 13457 134   
---------------------------------------------------------
||124579 3     12579 |2479  149    8     |459   6     249   
||8      2456  2569  |23469 3469   239   |1     2345  7     
||12479  12467 12679 |5     13469  12379 |3489  234   23489


We have an Exocet base r46c1 target r1c2 r7c3
so r1c2=1257 r7c3=1257
Then, no to hard to show

1&5r46c1 not valid ( nothing in r1c5)
5&7r46c1 not valid (nothing in r1c5=
1&2r46c1 not valid (conflicting in r7 digit 2)

we have now r4c1 <5>
and new strong links
1r46c1 = 2r4c1
5r6c1 = 7r46c1

this helps, but it's not enough.

May-be a new search for rank0 logic should be restarted??
but from the past I have doubts it would give anything.

Here the PM after the start

Code: Select all
||124579 1257  3     |249   459    6     |479   8     149   
||4579   4578  579   |1     34589  359   |2     347   6     
||6      1248  129   |23489 7      239   |349   134   5     
---------------------------------------------------------
||127    12567 4     |36789 135689 13579 |35678 12357 1238 
||3      12567 12567 |678   1568   4     |5678  9     128   
||157    9     8     |367   2      1357  |34567 13457 134   
---------------------------------------------------------
||124579 3     1257  |2479  149    8     |459   6     249   
||8      2456  2569  |23469 3469   239   |1     2345  7     
||12479  12467 12679 |5     13469  12379 |3489  234   23489



champagne
champagne
2017 Supporter
 
Posts: 7453
Joined: 02 August 2007
Location: France Brittany

Re: The hardest sudokus (new thread)

Postby ronk » Mon Nov 29, 2010 1:33 pm

champagne wrote:We have an Exocet base r46c1 target r1c2 r7c3
so r1c2=1257 r7c3=1257
Then, no to hard to show

1&5r46c1 not valid ( nothing in r1c5)
5&7r46c1 not valid (nothing in r1c5=
1&2r46c1 not valid (conflicting in r7 digit 2)

we have now r4c1 <5>

Thanks champagne. I had the r1c2=1257 and r7c3=1257 ... but nothing beyond that. Now, even with your hints above, I still can't prove r4c1<>5.
ronk
2012 Supporter
 
Posts: 4764
Joined: 02 November 2005
Location: Southeastern USA

Re: The hardest sudokus (new thread)

Postby champagne » Mon Nov 29, 2010 1:50 pm

ronk wrote:
champagne wrote:We have an Exocet base r46c1 target r1c2 r7c3
so r1c2=1257 r7c3=1257
Then, no to hard to show

1&5r46c1 not valid ( nothing in r1c5)
5&7r46c1 not valid (nothing in r1c5)
1&2r46c1 not valid (conflicting in r7 digit 2)

we have now r4c1 <5>

Thanks champagne. I had the r1c2=1257 and r7c3=1257 ... but nothing beyond that. Now, even with your hints above, I still can't prove r4c1<>5.



That's not to difficult and , if I remind correctly, this is something we already found in Golden Nugget

5 r46c1 can only be 5&1; 5&2; 5&7
as the others are not possible, it must be (if 5 is true) 5&2 . We can erase 5r4c1

champagne
champagne
2017 Supporter
 
Posts: 7453
Joined: 02 August 2007
Location: France Brittany

Re: The hardest sudokus (new thread)

Postby ttt » Mon Nov 29, 2010 6:43 pm

champagne wrote:
ronk wrote:
champagne wrote:We have an Exocet base r46c1 target r1c2 r7c3
so r1c2=1257 r7c3=1257
Then, no to hard to show

1&5r46c1 not valid ( nothing in r1c5)
5&7r46c1 not valid (nothing in r1c5)
1&2r46c1 not valid (conflicting in r7 digit 2)

we have now r4c1 <5>

Thanks champagne. I had the r1c2=1257 and r7c3=1257 ... but nothing beyond that. Now, even with your hints above, I still can't prove r4c1<>5.

That's not to difficult and , if I remind correctly, this is something we already found in Golden Nugget

5 r46c1 can only be 5&1; 5&2; 5&7
as the others are not possible, it must be (if 5 is true) 5&2 . We can erase 5r4c1

Wow..., I had spend over hour to study about that then see that: OK, r4c1<>5 but the presenting for that deduction seems too, too... complex so I think that we’ll have to study it more time... :D

ttt
ttt
 
Posts: 185
Joined: 20 October 2006
Location: vietnam

Re: The hardest sudokus (new thread)

Postby ttt » Tue Nov 30, 2010 5:46 am

champagne wrote:We can erase 5r4c1

Yes, I did study more and remind me Silver Plate :D
Code: Select all
003006080000100206600070005004000000300004090098020000030008060800000107000500000 11.5

eleven’s puzzle ER11.5: Raw Rank = 5 (linksets - sets)
17 Sets = {1R157 5R157 7R157 157C6 157C8 46N1}
22 Links = {157c1 1n2 5n57 7n3 1b34568 5b24569 7b34568}
2 AURs = (51)R64C61, (75)R64C81 AUR points {aur 5r4c1 1r6c6 7r6c8 }
1 Elimination --> r4c1<>5

Image

ttt
ttt
 
Posts: 185
Joined: 20 October 2006
Location: vietnam

Re: The hardest sudokus (new thread)

Postby champagne » Tue Nov 30, 2010 7:08 am

HI ttt

No surprise, using Allan model, many things can be done in once.
Such SLG's remain out of the scope of my solver. I stop at "nearly rank o logic"

I don't know if you had a look to "swampy soil", I forward here a "finding" of abi

9.....8.....7...6.....8.5.4.3.8.....4...2.9....1..6.3...41......6...7...5...4...2 swampy soil

In that puzzle, after the rank 0 logic (floors 1367) has been applied, we are there


Code: Select all
__ A___ B____ C____ |D____ E____ F____ |G____ H___ I_____
1||9___ 12457 23567 |23456 136__ 12345 |8____ 127_ 137___
2||1238 2458_ 258__ |7____ 1359_ 2459_ |123__ 6___ 139___
3||1367 127__ 2367_ |2369_ 8____ 1239_ |5____ 1279 4_____

4||267_ 3____ 259__ |8____ 1a579 459__ |12467 245_ 1567__
5||4___ 578__ 5678_ |3d5D_ 2____ 1A3D5 |9____ 1578 167___
6||278_ 2589_ 1____ |459__ 579__ 6____ |247__ 3___ 578___

7||2378 289__ 4____ |1____ 3569_ 2589_ |367__ 589_ 356789
8||1238 6____ 289__ |259__ 359__ 7____ |134__ 4589 13589_
9||5___ 1789_ 3789_ |369__ 4____ 389__ |1367_ 1789 2_____


not to hard to show here 1r4c5 = 5r5c46
so we have somehow a link between the floors "1" and "5"

abi makes the following remark
in boxes 1;5;9 the "5" belonging to the solution are on a diagonal.
They are paired with "4", giving a kind of "conditional elimination"

I don't know if anything new can get out of it, but it is worth being investigated

I have no free time to do it now.

champagne
champagne
2017 Supporter
 
Posts: 7453
Joined: 02 August 2007
Location: France Brittany

Re: The hardest sudokus (new thread)

Postby eleven » Tue Nov 30, 2010 8:32 am

Here is another one (11.8/11.8/10.7)

Code: Select all
 +-------+-------+-------+
 | 1 2 . | 3 . . | . . . |
 | 4 . . | . 5 . | . . . |
 | . . 6 | . . 1 | 7 . . |
 +-------+-------+-------+
 | 5 . . | 2 . . | . . . |
 | . 3 . | . 1 . | . 4 . |
 | . . 1 | . . 6 | 8 . . |
 +-------+-------+-------+
 | . . 9 | . . . | . 7 . |
 | . . . | 1 . . | 9 . . |
 | . . . | . . 9 | . 6 8 |
 +-------+-------+-------+
        First Peak
Last edited by eleven on Tue Nov 30, 2010 12:39 pm, edited 1 time in total.
eleven
 
Posts: 3142
Joined: 10 February 2008

Re: The hardest sudokus (new thread)

Postby champagne » Tue Nov 30, 2010 9:01 am

eleven wrote:Here is another one (11.8/11.8/10.7)



Good catch; that one is closer to "golden Nugget" "Silver Plate".

Same magnitude of difficulty without special tool
. print over 600K
. necessity to pass AAHS 2 cells
. nested derived links level 5

An efficient EXOCET in r8c89 reducing significantly the final complexity (final print below 200K, but same level of nested devrived weak links)

for "ttt", there is a potential for a nice SLG using floors 12345, but i did not find a "rank 0" logic

is worth a name

champagne
champagne
2017 Supporter
 
Posts: 7453
Joined: 02 August 2007
Location: France Brittany

Re: The hardest sudokus (new thread)

Postby eleven » Tue Nov 30, 2010 12:38 pm

Thanks again,
so i will call it "First Peak" in the hope, that there are more ones.

This one has ER 11.7/11.7/10.7
Code: Select all
 +-------+-------+-------+
 | 1 2 . | 3 . . | . . . |
 | 3 . . | . 4 . | . . . |
 | . . 5 | . . 2 | 3 . . |
 +-------+-------+-------+
 | 5 . . | . . 6 | 1 . . |
 | . . . | . . . | . 9 . |
 | . . 6 | 7 . . | . . 8 |
 +-------+-------+-------+
 | . 5 . | . . 1 | 6 . . |
 | . . . | 9 . . | . 4 . |
 | . . . | . 8 . | . . 7 |
 +-------+-------+-------+

[edit: removed a redundant clue - guess changes the ER to 11.7/1.2/1.2]
eleven
 
Posts: 3142
Joined: 10 February 2008

Re: The hardest sudokus (new thread)

Postby champagne » Tue Nov 30, 2010 3:29 pm

That one is not too difficult

(print slightly below 200K and solved just using AAHS 2 cells)

An EXOCET in r12c3 push down the difficulty in a reasonable area (print about 130K)
No rank 0 logic, but a promising 5 floors situation to find a big SLG (EXOCET digits plus digit '5')

would be in the low part of my hardest file

I am preparing a lot of puzzles coming out of the last game with much lower SE rating but seen as tougher by my solver

champagne
champagne
2017 Supporter
 
Posts: 7453
Joined: 02 August 2007
Location: France Brittany

Re: The hardest sudokus (new thread)

Postby ttt » Tue Nov 30, 2010 4:58 pm

champagne wrote:
eleven wrote:Here is another one (11.8/11.8/10.7)

..........
is worth a name

Yes, they should be named... :D
I'll study this one and present for you if I would be found something news :D
Wow..., one bottle vodka is enough for me.. :oops:

ttt
Last edited by ttt on Tue Nov 30, 2010 5:06 pm, edited 1 time in total.
ttt
 
Posts: 185
Joined: 20 October 2006
Location: vietnam

Re: The hardest sudokus (new thread)

Postby eleven » Tue Nov 30, 2010 5:03 pm

champagne wrote:I am preparing a lot of puzzles coming out of the last game with much lower SE rating but seen as tougher by my solver

It seems that i can find top puzzles for each public rating now (at least for so different ones like ER and q2).

Since i respect your rating most, a public version of your rating program would be very appreciated. Any plans for that ?
(Otherwise we have little chances to get into your top list)
eleven
 
Posts: 3142
Joined: 10 February 2008

Re: The hardest sudokus (new thread)

Postby champagne » Tue Nov 30, 2010 5:35 pm

eleven wrote:
champagne wrote:I am preparing a lot of puzzles coming out of the last game with much lower SE rating but seen as tougher by my solver

It seems that i can find top puzzles for each public rating now (at least for so different ones like ER and q2).

Since I respect your rating most, a public version of your rating program would be very appreciated. Any plans for that ?
(Otherwise we have little chances to get into your top list)


Up to now, nobody expressed the willingness to use it, so I did not do any cleaning work to go in that way, but I can give it to anybody, that's not a problem.

It would easier now that i am using the free microsoft visual c++ platform to compile it.

In fact i have only one program to solve, generate, work out a canonical form ..., so I would have at least to comment on the options panels

If you want to go deeper, we can start thru PM or email.


champagne
champagne
2017 Supporter
 
Posts: 7453
Joined: 02 August 2007
Location: France Brittany

PreviousNext

Return to General