The New Sudoku Players' Forum

by **m_b_metcalf** » Fri Aug 25, 2017 4:49 pm

hkociemba wrote: I wonder what causes these differences between our opinions.

Just crossed-wires: by redundant I mean wrt to any solution method. Sorry for any confusion.

by **hkociemba** » Sat Aug 26, 2017 5:59 pm

m_b_metcalf wrote: I could solve all the puzzles, but my impression is that they are harder (require more elaborate techniques) than you state in your comments.

Maybe we also should try to track down this issue. The first one (16x16, 93 givens) )is really solvable with pairs only, I append the verbose output, it is not too large and maybe you can see where you fail in the solving process.

Hidden Text: Show

by **hkociemba** » Sat Aug 26, 2017 7:31 pm

After quite a long search I now also found an interesting 9x9 example.

.5......2..15..968..........1...5...8..12......568.....4...91...72..345......6..7

Hidden Text: Show

is pair-optimal and removing the 5 in r6c3 it gets triple-optimal and also overall optimal:

.5......2..15..968..........1...5...8..12.......68.....4...91...72..345......6..7

Hidden Text: Show

I hope Mike that this helps to track down the problems.

by **m_b_metcalf** » Sun Aug 27, 2017 10:37 am

m_b_metcalf wrote: In fact, it would perhaps help with my diagnostics if you could make a few 9x9 and 16x16 puzzles available.

Thanks for all those puzzles. I have now identified one silly reason that I thought that these were harder than you stated, namely that in my testing code I'd left a section activated that should not be. When I correct that, we agree on all the 9x9, 16x16 and 25x25 puzzles. That leaves the original problem I had with, for instance, your 144x144_triples.txt, and I will work on that next week. Sorry for any inconvenience.

Regards,

Mike

P.S. I had a bit of a problem with your terminology. You appear to use the term 'block' for a technique for what we commonly call pointing, as well as also for a region of a puzzle that we commonly call a box. Also, we usually call the attribute that you refer to as 'optimal' as minimal.

by **Pat** » Sun Aug 27, 2017 10:45 am

m_b_metcalf wrote:

You appear to use the term 'block' for a technique for what we commonly call pointing,
as well as also for a region of a puzzle that we commonly call a box.

no, he uses block for box
-- that's fairly common

b\r

by **m_b_metcalf** » Sun Aug 27, 2017 10:56 am

Pat wrote:block_candidates_in_row"
is b\r
(the candidates for S in box B
are all in row R)[/list]

... which is pointing (according to SE) rather than pairs or triplets.

by **hkociemba** » Sun Aug 27, 2017 11:12 am

Since I deal with Sudoku just since a few month any comments from experienced "Sudokuers" about terminology are welcome. I did not like the term "pointing" and "claiming" because they give me no clue what they really mean. So I decided to use block_candidates_in_row and row_candidates_in_block for example. The first one is called pointing, the second is called "claiming" acoording to http://hodoku.sourceforge.net/en/tech_intersections.php. Eventually the second is called also "pointing" by some people. Of course it would be no problem to add the terminology in my program in brackets if this would reduce confusion.
I will use "minimal" instead of "optimal" in future and eventually edit my older posts.

by **m_b_metcalf** » Fri Sep 01, 2017 12:38 pm

m_b_metcalf wrote:..., we agree on all the 9x9, 16x16 and 25x25 puzzles. That leaves the original problem I had with, for instance, your 144x144_triples.txt, and I will work on that next week. Sorry for any inconvenience.

I am now happy to report that new_100x100_5848__tripleOptimal now also yields to my solver, with its new hidden triplets code. However, it does not solve 144x144_triples, and before I investigate further could you please confirm that it requires only basic methods up to triples, just like the 100x100.

Thanks,

Mike Metcalf

by **hkociemba** » Sat Sep 02, 2017 7:45 am

Yes I can confirm this, I reloaded the file from this forum to go sure that we use the same file. The solution is appended. It uses 131 times hidden triples and 112 times naked triples, the first time a triple is needed to continue the solving proceess is line 671. Btw. I still run the reduce program on the 225x225 and meanwhile there are only 35514 givens left. About 20% of the cells have been checked now...

by **m_b_metcalf** » Sat Sep 02, 2017 8:57 am

hkociemba wrote:Yes I can confirm this, I reloaded the file from this forum to go sure that we use the same file. The solution is appended. It uses 131 times hidden triples and 112 times naked triples, the first time a triple is needed to continue the solving proceess is line 671.

Thanks. However, by the time the program stops it has performed 368 naked (candidate) eliminations and 165 hidden. There is an effect known from 9x9s whereby different solution paths can very occasionally have different maximum degrees of difficulty, and perhaps that is occurring in such monsters. I'll investigate further when I have time.

Regards,

Mike Metcalf

by **hkociemba** » Sat Sep 02, 2017 12:59 pm

m_b_metcalf wrote:There is an effect known from 9x9s whereby different solution paths can very occasionally have different maximum degrees of difficulty

Thats really new for me. When does that happen? If I allow for example only singles it seems quite obvious for me that I always end up at the same point, no matter in which order I proceed. So which methods could have such an effect?

by **m_b_metcalf** » Sat Sep 02, 2017 1:18 pm

hkociemba wrote:
m_b_metcalf wrote:There is an effect known from 9x9s whereby different solution paths can very occasionally have different maximum degrees of difficulty

Thats really new for me. When does that happen? If I allow for example only singles it seems quite obvious for me that I always end up at the same point, no matter in which order I proceed. So which methods could have such an effect?

I'll have to rely on pat's elephantine memory to find an answer to that. It was a long time ago, and maybe on the Programmers' Forum. Obviously, it was more complicated than just singles.

by **Pat** » Sun Sep 03, 2017 7:46 am

m_b_metcalf wrote:There is an effect known from 9x9s
whereby different solution paths
can very occasionally have different maximum degrees of difficulty

sorry but the ONLY case i know
involves deducing-from-uniqueness-of-answer.

even then,
one merely needs to remember the difference
between solved cells and given cells,
and the "uniqueness" argument re-appears.

by **Pat** » Sun Sep 03, 2017 7:47 am

m_b_metcalf wrote:
hkociemba wrote:uses 131 times hidden triples
and 112 times naked triples

the first time a triple is needed is line 671

by the time the program stops
it has performed 368 naked (candidate) eliminations
and 165 hidden.

for diagnostics,
could you please post the known cells
at the point that your program stops?

hkociemba

by **m_b_metcalf** » Sun Sep 03, 2017 8:32 am

Pat wrote:for diagnostics,
could you please post the known cells
at the point that your program stops?

then hkociemba could run his program
starting at that point---

That's a good idea that I'll come back to if my investigation gets no further.

The New Sudoku Players' Forum

giant sudoku's (16x16, 25x25, 36x36 .... 100x100)

Re: giant sudoku's (16x16, 25x25, 36x36 .... 100x100)

Re: giant sudoku's (16x16, 25x25, 36x36 .... 100x100)

Re: giant sudoku's (16x16, 25x25, 36x36 .... 100x100)

Re: giant sudoku's (16x16, 25x25, 36x36 .... 100x100)

re: terminology

Re: re: terminology

Re: giant sudoku's (16x16, 25x25, 36x36 .... 100x100)

Re: giant sudoku's (16x16, 25x25, 36x36 .... 100x100)

Re: giant sudoku's (16x16, 25x25, 36x36 .... 100x100)

Re: giant sudoku's (16x16, 25x25, 36x36 .... 100x100)

Re: giant sudoku's (16x16, 25x25, 36x36 .... 100x100)

Re: giant sudoku's (16x16, 25x25, 36x36 .... 100x100)

re: 144-squared triples

re: 144-squared triples

Re: re: 144-squared triples