The New Sudoku Players' Forum

by **Pat** » Mon Apr 02, 2012 11:16 am

champagne wrote:
compared to games with (slightly) smaller number of clues,
I generated a relatively low number of puzzles with the pattern of the game 170

you mean, few minimal puzzles

by **champagne** » Mon Apr 02, 2012 11:23 am

Pat wrote:
champagne wrote:
compared to games with (slightly) smaller number of clues,
I generated a relatively low number of puzzles with the pattern of the game 170

you mean, few minimal puzzles

higher number of clues ==> greater chance the puzzle is non-minimal

may be you have the reason.

Due to the rules in the pattern game, non minimal puzzles are filtered in my process

but it would be interesting to see if other players got similar results

champagne

by **champagne** » Thu Jul 19, 2012 6:56 am

few comments on the last game.

I generated about 50 millions puzzles rating 4.5 and more.
I did not finish the serate rating, but I think I would not have found the last "more" of Joël.

More interesting is the spread of ratings ED9.9 and more.

I several times tried to see whether it was possible to establish a strategy based on the probable pattern of such puzzles, but I failed, likely due to the fact that the scan search remains very very low compared to the vicinity search.

the list of puzzles is sorted on the minlex form delivered by gridchecker.

As in nearly all the cases I studied, no puzzle starts with digit 1234567 in that order.
And as often, many puzzles have a duplicated digit in the first 5

Hidden Text: Show

by **champagne** » Tue Aug 14, 2012 3:57 pm

game 183 is now closed.

I am not sure anybody is interested in explanations on why the gap was so big in that game.

Anyway, I intended to make public the code I use and the general process I follow in such games.
I started to go in that direction and opened a new project on google code area

project skmpp

champagne

by **m_b_metcalf** » Wed Aug 15, 2012 8:39 am

champagne wrote:I am not sure anybody is interested in explanations on why the gap was so big in that game.

champagne,
Alas, that is probably true. It really does seem that the intensive sudoku studies that took place here and on the Programmers' Forum, from the early days in 2005 and onwards, have virtually ground to a halt. (There has been no substantive posting on the Programmers' site for months.) What we need now is someone to write a record of everything that has been done and achieved so that it can be conveniently accessed by posterity. I once asked for volunteers to write a book (I have contacts to a leading publisher who was interested) but the only candidate withdrew for family reasons.

Sigh,

Mike Metcalf

by **champagne** » Tue Aug 28, 2012 5:51 pm

m_b_metcalf wrote: Alas, that is probably true. It really does seem that the intensive sudoku studies that took place here and on the Programmers' Forum, from the early days in 2005 and onwards, have virtually ground to a halt. (There has been no substantive posting on the Programmers' site for months.) What we need now is someone to write a record of everything that has been done and achieved so that it can be conveniently accessed by posterity. I once asked for volunteers to write a book (I have contacts to a leading publisher who was interested) but the only candidate withdrew for family reasons.

Sigh,

Mike Metcalf

Hi Mike,

Despite the lack of common interest, I continue to use that pattern game as testing platform to evaluate my investigation process.

The current game (185) is not very active, but we are still in the summer holidays season.
I am running that test using exclusively public tools described in the project pointed 2 posts above.

We are close to the end of day 3 and my database has about 100 millions puzzles rating 4.5 and more. This has been done using the full power of an i7, expanding in blind mode most of the puzzles diamonds 8.8 and more (+-3) .

After one hour, I had in cache about 15 diamonds (some of them still there) and 2 10.x/1.2/1.2 puzzles. This came looking for symmetry of given

I am trying to cover all the field and at that point, I have nearly 30 valid submissions in cache, including self trumping. I am missing rare ratings as 5.1; 6.1

Interesting point, in line with previous games, it took several days to focus on the best areas. I found some hours ago an ED 9.7 (self trumping ) but isolated.

interesting pattern anyway. Thank you for having proposed it to the game.

Best regards

champagne

by **ronk** » Tue Aug 28, 2012 9:26 pm

champagne wrote:I am running that test using exclusively public tools described in the project pointed 2 posts above

I see nothing "public" in the project you cite.

by **champagne** » Tue Aug 28, 2012 9:38 pm

ronk wrote:
champagne wrote:I am running that test using exclusively public tools described in the project pointed 2 posts above

I see nothing "public" in the project you cite.

you have the sources, you have the projects definitions, what more is needed to be public ???

I have still some documentation to write, and some .bat files to describe, but a significant part of the doc is there

when and if anybody shows interest in that, I'll surely prepare an upload with the .exe, as we did for skfr, but we are far below that point

by **ronk** » Tue Aug 28, 2012 10:31 pm

champagne wrote:
ronk wrote:
champagne wrote:I am running that test using exclusively public tools described in the project pointed 2 posts above
I see nothing "public" in the project you cite.
you have the sources, you have the projects definitions, what more is needed to be public ???

How about *.cpp, *.h, etc. files in a *.zip or other compressed format so that people don't have to be familiar with the "svn" tool? [I think we had a very similar discussion for skfr.] Otherwise, how about hints on how to obtain and use our own copies of "svn"?

BTW how about makefiles for the compiler(s) supported?

by **m_b_metcalf** » Wed Aug 29, 2012 9:46 am

champagne wrote:interesting pattern anyway. Thank you for having proposed it to the game.

champagne,
I'm glad this pattern was useful.

I have now stopped all program development, and use my existing codes with various options during a game. One that is played slowly suits me better as I have limited resources (and time). I'm happy to get a decent result, winning is impossible. In the foreseeable future we'll run out of patterns, and then it will be time for someone to write an article about the game.

Regards,

Mike

by **champagne** » Wed Aug 29, 2012 10:19 am

m_b_metcalf wrote: In the foreseeable future we'll run out of patterns, and then it will be time for someone to write an article about the game.
Regards,
Mike

I don't think there is a big risk of running out of patterns, the 21 clues field is surely covered, but we have still a lot of room in patterns with more clues, but surely, the symmetry constraint, the easiest to change brings severe limitations (if we want interesting patterns)

For me, the game is already dead by lack of players.

I am more thinking of exploring patterns to get a better knowledge of the sudoku field. I continue my lonesome search of "potential hardest" puzzles, exploring to-day the 25 clues area.

And on the solving side, I am still working on building functions to crack such puzzles with as much elegance as possible.

As I already wrote, the pattern games remains a terrific place to evaluate the performance of the software used. I'll continue to participate with a limited use of my findings as long as the game will continue

Best regards

champagne

by **m_b_metcalf** » Wed Aug 29, 2012 11:17 am

champagne wrote:I don't think there is a big risk of running out of patterns,

I meant not the existence of patterns but the fact that hardly any more are being submitted. It's actually quite a lot of work to find a 'good' one, and the lack of players means there's little incentive to develop them.(Notice in the current pattern that eight of the boxes have permuted diagonals, some with an additional or missing cell.)

All good things come to an end.

Regards,

Mike

by **champagne** » Fri Nov 23, 2012 10:03 am

revising my new code to bring in the old process of "symmetry of given", I went trough the published files

potential hardest
sample file of 7 million puzzles.

looking first for symmetries fitting with the pattern game constraints.

I have seen many patterns that could be used in the game with good chances to be of interest.

One condition to have fun exploring such patterns would be to stick strictly to the game basic rule "start from scratch".

I made an exception in the last game mainly because I could not imagine to be an active player starting in day 5. I re used the known puzzle from eleven.

If other players want to go in that direction, I can look for a valid start and submit such patterns to gremlin.

by **m_b_metcalf** » Sun Nov 25, 2012 12:39 pm

champagne wrote:I have seen many patterns that could be used in the game with good chances to be of interest.

One condition to have fun exploring such patterns would be to stick strictly to the game basic rule "start from scratch".

I made an exception in the last game mainly because I could not imagine to be an active player starting in day 5. I re used the known puzzle from eleven.

If other players want to go in that direction, I can look for a valid start and submit such patterns to gremlin.

Seems a good way to go. I think that there should then just be an understanding that the dealer doesn't play the initial seed for, say, 48 hours, nor use it himself as a seed until after it's posted.

Bon dimanche,

Mike

by **champagne** » Tue Nov 19, 2013 10:00 am

WosRet wrote:::: comment :::

champagne wrote:all diamonds from 8.2 to 9.7 except 9.6

Impressive! I think, I have to change my "search system"...

some comments on game 216 (current game)

I tested that game thinking that blue would come again, playing on the high side as in game 215, but he did not show up (I'll do a separate comment on game 215)

In that game, I applied the same strategy as in former games

opening the game looking for seed through a deep search on the initial seed
adding seeds using the symmetry of given (nearly nothing in that case)

then vicinity +-3 with some deeper search on the highest ratings having high ED's

I generated more than 160 million puzzles.

As usual, I filter first the files using my fast classification process.

In that pattern where a very small number of puzzles passed the dynamic serate process, I used "group diamonds" as new seed. (group diamond is what appears as diamond in my fast classification process, skipping the serate search for the shortest path)

I used skfr as second filter to locate puzzles of potential interest and I had to rate with serate for a very small number of puzzles.

Moreover, ER 8.0 and 8.1 came through "group pearls" so I could ignore all other puzzles with a fast rating over 6.1.

In total, I used about 50% of an i7 based computer over 5 days

The New Sudoku Players' Forum

Patterns Game Strategies

Re:

Re: Patterns Game Strategies

Re: Patterns Game Strategies

Re: Patterns Game Strategies

Re: Patterns Game Strategies

Re: Patterns Game Strategies

Re: Patterns Game Strategies

Re: Patterns Game Strategies

Re: Patterns Game Strategies

Re: Patterns Game Strategies

Re: Patterns Game Strategies

Re: Patterns Game Strategies

Re: Patterns Game Strategies

Re: Patterns Game Strategies