scan solution grids for 17 clues as of blue

Programs which generate, solve, and analyze Sudoku puzzles

Re: scan solution grids for 17 clues as of blue

Postby champagne » Tue Dec 21, 2021 9:04 am

Ruud from Sudocue wrote:Since the list assembled by Gordon Royle has been hidden behind a (misconfigured?) firewall, it has become difficult to find a single source for a complete list of 17-clue sudokus.
A few sources have a list of 49158 puzzles and there are several references to this thread.
If this is not the place to post new additions, please point me to the correct thread.

I would like to contribute the following list of 556 puzzles, which, to my best knowledge, are not part of the current 49158.
All of these have been found by replacing a single clue from one of the original 49518.
To make sure that these are indeed new additions, please verify them to be unique, correct and in the right canonical form.


Hi Ruud,

so many new 17 would have been amazing. We are expecting at most one or 2 more 17 (if we are lucky).

Nothing new on my side. However, as I am using a data base in max lexical canonical form, it's not easy for me to give you the original corresponding 17.

EDIT AFAIK, mladen dobrichev made a full vicinity search "deep 3". Such puzzles would have been seen in the search.
champagne
2017 Supporter
 
Posts: 7257
Joined: 02 August 2007
Location: France Brittany

Re: scan solution grids for 17 clues as of blue

Postby Ruud from Sudocue » Tue Dec 21, 2021 12:53 pm

Thanks champagne,
I fixed the error in my min-lex algorithm. It no longer produces any additions.
Ruud from Sudocue
 
Posts: 14
Joined: 10 December 2021

Re: milestone December 2021

Postby champagne » Tue Feb 08, 2022 8:54 am

milestone 2022/02/08 last one 2021/12/03

A little more than the usual 2 months between milestones.

We have only 23 more "bands 1" closed. This is likely a good indicator of the average expected run time for the remaining last 62 "bands 1" to cover.
85% of the "bands 1" have been processed, but we have still about 25% of the solution grids to scan.
No bet at this point about the probability to find a new 17.

Our friend mith joined the team and checked his first grid (index 120). Still busy to produce high ratings, he supplied a limited power to the 17 search, but this could overcome the increase in the average run time and give us a chance to finish the job around mid 2022

all bands 1 not processed are in the index range 63-129.
Mathimagics is processing the slice 110_119
Mith is processing the slice 120-129
and I am working in the slice 63-70

Another new, but the final result will come to late to help us in the current search.

Clearly, blue continued to work on this topic and supplied mathimagics with a DLL searching the 18 with amazing low run times. This has been discussed in another thread
low_hi clue
Without the sources, just with few comments given here and there, it was difficult to understand what he did new, but analyzing the results, some points appear.

This could be the basis for a new attack of the 17 search giving much better results, but the current scan will be over before the delivery of a fully tested fresh code.
However, this could be of help to do start a 18 search in a similar way

The first step for this new code is to finish and test the implementation of a "one solution grid" scan. Where I am, this should be over in February.
champagne
2017 Supporter
 
Posts: 7257
Joined: 02 August 2007
Location: France Brittany

Re: milestone December 2021

Postby champagne » Sun Apr 10, 2022 7:29 am

milestone 022/04/10 after 2022/02/08

Nothing really new, may be a remark : the last bands are slightly harder than expected.

Mith, in trouble with his computer had to stop a while his contribution.
Mathimagics and me continue to bring the same power to the scan.

We still have 41 "bands 1" to scan to likely establish that all 17 clues are known, but we can not exclude the possibility to produce the #49159!!

Blue had written a 18/19 search much more efficient that the current scan. Unhappily, we don't have the sources and the DLL in the hands of mathimagics is for a solution grid, not for all bands 3 attached to a given pair band1;band2.

I am still fighting to crack blue's findings, but not yet in line with his results.
The scan will be over in 3/4 months, so a new version will come to late to replace the current one.
I'll continue the exercise to try to deliver sources performing well in the search of low clues linked to a solution grid
champagne
2017 Supporter
 
Posts: 7257
Joined: 02 August 2007
Location: France Brittany

Previous

Return to Software