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 mith » Wed Sep 06, 2023 5:01 pm

IIRC every puzzle in the te3 db I have checked with uniqueness turned off has SER 11.7+. I should run this on a less biased sample though (if not re-running all 1 million puzzles).
mith
 
Posts: 996
Joined: 14 July 2020

Re: The hardest sudokus (new thread)

Postby eleven » Wed Sep 06, 2023 7:12 pm

Just to say it, before uninformed users think, that Paquita's T&E(3) puzzles are hard:
All posted here have a (easy-to-spot) single extra candidate of a tridagon pattern, which must be true. Then the puzzles are rather easy, none harder than SER 7.x.

[Added:]If someone thinks, she can solve the hardest, try those in champagne's database (there), ph_2010.zip. (i cannot go there for 2 days, because i refused to confirm, that google knows my handy ;) ).
eleven
 
Posts: 3150
Joined: 10 February 2008

Re: The hardest sudokus (new thread)

Postby Paquita » Wed Sep 06, 2023 7:47 pm

This is the link to the updated T&E(2) collection https://drive.google.com/file/d/1BZwwZbpW-lfj6xMtzR-OW2iEMIQaqAgs/view?usp=sharing
It is ph2010 with newer puzzles from mith, hendrik, and some from JPF and me.
I included the original forms of the puzzles, after the maxlex form.

Is it generally true that T&E(3) puzzles are not hard?

Mith has done such a lot to collect T&E(3) puzzles that I thought I would focus on T&E(2), it is now a much smaller collection. But given that I have sometimes used T&E(3) puzzles as seeds, I do end up with some T&E(3) puzzles as well. I will post them seperately.
Paquita
 
Posts: 132
Joined: 11 November 2018

Re: The hardest sudokus (new thread)

Postby eleven » Wed Sep 06, 2023 7:55 pm

Paquita wrote:Is it generally true that T&E(3) puzzles are not hard?

No, but a good part of them, and the diffculty seems to correlate with the number of extra canditates of the tridagon pattern.
eleven
 
Posts: 3150
Joined: 10 February 2008

Re: The hardest sudokus (new thread)

Postby denis_berthier » Thu Sep 07, 2023 2:58 am

Paquita wrote:This is the link to the updated T&E(2) collection https://drive.google.com/file/d/1BZwwZbpW-lfj6xMtzR-OW2iEMIQaqAgs/view?usp=sharing
It is ph2010 with newer puzzles from mith, hendrik, and some from JPF and me.
I included the original forms of the puzzles, after the maxlex form.

I'm sure this will be very helpful to mith when he starts dealing with the T&E(2) collection. I'll also start to have a look at it.

Paquita wrote:Is it generally true that T&E(3) puzzles are not hard?

No. You can ask the same question about puzzles containing a full quad (4 candidates in each of the 4 cells). Such puzzles may be in T&E(2). But if you know the Quad pattern, they become much easier to solve, in general.
All the known patterns in T&E(3) have a tridagon with more or less guardians. If you know the tridagon pattern and rules able to use it, you can simplify the puzzle. In most cases, it doesn't become easy at all.
The existence of T&E(3) puzzles and the use of the T&E-depth criterion have been a breakthrough in the search of "hard" puzzles. They raise questions of their own that SER is unable to tackle.

Paquita wrote:Mith has done such a lot to collect T&E(3) puzzles that I thought I would focus on T&E(2), it is now a much smaller collection. But given that I have sometimes used T&E(3) puzzles as seeds, I do end up with some T&E(3) puzzles as well. I will post them seperately.

It has become clear that the two collections have separate search criteria and separate goals: T&E-depth vs highest SER (or PGX).
None of the two collections can claim to have the goal of finding "the hardest puzzles", because this expression is meaningless if you don't say hardest wrt what.
The T&E(2) collection is currently smaller because mith's technique of expansion hasn't been systematically applied to it.


After a closer look, there's still something wrong in your collection: it lacks a uniform presentation.
- you should use either a semi-colon or a space as separator, but not both. (semi-colon is much better IMO).
- for ER EP ED, you should use either 3 fields ER;EP;ED or one ER/EP/ED, but the choice should be the same everywhere (empty fields would appear as ;; )
- for ER, EP or ED, use the same notation everywhere: either 11.9 or 11.90, not both.
- a first line should name the fields, e.g.: puzzle; orig-puzzle; ER; EP; ED; collection, creator; name; nb_in_collection; ?;
(what does "1;22;" or "11523;23;" means at the end of the 1st two lines?

.
denis_berthier
2010 Supporter
 
Posts: 4190
Joined: 19 June 2007
Location: Paris

Re: The hardest sudokus (new thread)

Postby Paquita » Thu Sep 07, 2023 7:18 am

Denis,

I know my collection lacks uniformity. I collected puzzles from various sources and the owners of it used different formats for the rating. I have not tried to put them all in the same format, rather I just copied their files. Some fields are not available in every entry and some use space instead of semicolon.
1;22 comes from the ph2010 database, there it is puzzle number 1 with 22 givens. But I did not number or name the puzzles and did not count the number of clues.

I know the current file still needs work to become a neat database of puzzles. It suffices for my needs as it is now. I thought mith would publish a follow-up on ph2010 and this file might help. But, well, I suppose I could edit this file myself, if mith is okay with it.
Paquita
 
Posts: 132
Joined: 11 November 2018

Re: The hardest sudokus (new thread)

Postby denis_berthier » Thu Sep 07, 2023 8:43 am

Paquita wrote:I know the current file still needs work to become a neat database of puzzles. It suffices for my needs as it is now. I thought mith would publish a follow-up on ph2010 and this file might help. But, well, I suppose I could edit this file myself, if mith is okay with it.

Did you try to eliminate the T&E(3) puzzles?
The following 11.8:
Code: Select all
9876.........958.......4...7.8......36.2..7...293.768.2.6....97..3...2.8......36.;9876.........958.......4...7.8......36.2..7...293.768.2.6....97..3...2.8......36.   11.8/1.2/1.2 monh

is in T&E(3).

I'm currently checking the 11.9s and 11.8s in your list for possibly new B7Bs (or for any unexpected B8B). After the publication of the long lists of T&E(3) puzzles, I had stopped checking all the T&E(2)s.
.
denis_berthier
2010 Supporter
 
Posts: 4190
Joined: 19 June 2007
Location: Paris

Re: The hardest sudokus (new thread)

Postby Paquita » Thu Sep 07, 2023 9:25 am

Denis,

I only checked my own puzzles for T&E (puzzles with attribute PAQ). As for mith and Hendriks puzzles, I totally relied on miths collection of T&E(3), and removed the puzzles that are in that collection. I am surprised that a T&E(3) from Hendrik is still in this file, I suppose it means that it is not in miths T&E(3) collection. It would take me a long time to check all puzzles for T&E myself.
Paquita
 
Posts: 132
Joined: 11 November 2018

Re: The hardest sudokus (new thread)

Postby hendrik_monard » Thu Sep 07, 2023 10:40 am

I compared the TE2 database posted here yesterday by Paquita with my local database. I started with the oldest part, containing the 1161 puzzles from the file 'all116-20211103' posted by mith in November 2021. Of them, all but the following two puzzles were present in Paquita's database:

98.7.....6...5.8....4....3.7..9..5....3....4......2..1.6...1..2..9.........89.6..;22;11.8;1.2;1.2;0;95732;hendrik_monard
98.76....7.....9...54......6...3.8....5....2......4.1..7.38.......9..3.......1..2;22;11.7;11.7;2.6;97235;98806;mith

The original puzzles may have been an isomorph of these.
I guess that they didn't survive the TE2 selection process. However, based on the result from my solver, I find it very difficult to believe that these 2 puzzles would be in TE3. Could someone confirm that?

Edited: number of puzzles 1161 (not 535)
Last edited by hendrik_monard on Thu Sep 07, 2023 1:23 pm, edited 1 time in total.
hendrik_monard
 
Posts: 96
Joined: 19 April 2021
Location: Leuven (Louvain) Belgium

Re: The hardest sudokus (new thread)

Postby Paquita » Thu Sep 07, 2023 11:54 am

Hendrik,

They are in T&E(2). I don't know how they got lost. I will add them.


anyone,
I have no experience with gridchecker and I can't find documentation. Can someone give me the commandline arguments for : input "myfile.txt", output : minimals of the puzzles in that file?
Paquita
 
Posts: 132
Joined: 11 November 2018

Re: The hardest sudokus (new thread)

Postby hendrik_monard » Thu Sep 07, 2023 1:38 pm

Paquita wrote:
I have no experience with gridchecker and I can't find documentation. Can someone give me the commandline arguments for : input "myfile.txt", output : minimals of the puzzles in that file?


I use Glen Fowler's sudoku solver / generator with he following commandline :
sudoku-64.exe -qFN -f%#mc puzzles.txt > minlex.txt
The programme can be downloaded from: http://gsf.cococlyde.org/download/sudoku
hendrik_monard
 
Posts: 96
Joined: 19 April 2021
Location: Leuven (Louvain) Belgium

Re: The hardest sudokus (new thread)

Postby Paquita » Thu Sep 07, 2023 1:50 pm

Thank you hendrik, I have gsf, but I need not minlex, but minimals. I have been told that gridchecker can do that but I can't figure out what command line it expects.
Paquita
 
Posts: 132
Joined: 11 November 2018


Postby 1to9only » Thu Sep 07, 2023 2:58 pm

Paquita wrote:I have no experience with gridchecker and I can't find documentation. Can someone give me the commandline arguments for : input "myfile.txt", output : minimals of the puzzles in that file?

Code: Select all
gridchecker.exe --help
gridchecker.exe --solve --minimals < puzzles.txt > output.txt
grep.exe -v redundant output.txt > minimal.txt
User avatar
1to9only
 
Posts: 4177
Joined: 04 April 2018

Re: The hardest sudokus (new thread)

Postby Paquita » Thu Sep 07, 2023 3:07 pm

I tried

gridchecker --scan --gridlist <inputfile> outputfile

where the inputfile is in brackets, the outputfile not. It produces an empty outputfile so something is right.
But this seems not to be the command options for minimal puzzles.
Paquita
 
Posts: 132
Joined: 11 November 2018

PreviousNext

Return to General