The New Sudoku Players' Forum

[StrmCkr]

Kira : welcome to thr forums

The 2nd pattern is correct despite what you think

N cells with n +1 digits
Not all cells need to have each of n candidates ad long as the overal set found has exactly n+1 digit in n cells.

4678 in 3 cells.

Most solvers typically ignore naked subsets when searching, but they are still valid.

[Cenoman]

Release 608

YZF_Sudoku608.part1.rar

YZF_Sudoku608.part2.rar

Sorry for a very naive question: what are the different files "part1" and "part2" for ?
I have tried to download both (under Windows 10 OS) Part2 seems to contain the same filenames as part1 but is unzipped with errors. Which instruction have I missed ?
Thanks in advance.

[yzfwsf]

Because the forum limits the capacity of a single attachment, compressed files are divided into volumes. Therefore, you must download two files at the same time, put them in the same folder, and then unzip them.

[ghfick]

Here are a few more suggestions and queries:

May I suggest adding the SK Loop? It is true that the SK Loop can overlap considerably with MSLS. pjb has noted that there are puzzles where the two steps can give different exclusions though. The logic is quite different though.
While using 'Show Next Step', one can still see Forcing Chains before seeing MSLS/JE.
With your implementation of JE, there are the interesting cases where additional cells can be added to the "S" cells to confirm the JE pattern. Currently the solver says: ' "S" Cells Need Include xxxx True Base Cands xxxx'. I think this needs rewording. The additional cells are not "S" cells and there are no known True Base Candidates [I am unsure what is intended here?]
Since there are Naked Quads and Hidden Quads, can there be Almost Locked Quads? Maybe such an ALQ step is getting too close to Sue de Coq or ALS-XZ? The logic would be different though.
One receives the Avoidable Rectangle in solution paths. It should listed in 'Techniques'. Recent forum discussion on the AR may be relevant. [UR 1.1]. Alas, I understand very little of that discussion.
I am unsure of the boundaries between the levels of difficulty. I think your intention was to use equivalents to the default boundaries in HoDoKu but now using the original SE ratings. What is the boundariy between Extreme and Insane?
SE contains only a portion of the steps now in your solver. So sometimes, an MSLS or a JE takes an 'Insane' puzzle down to a lower level [like 'Unfair']. Nevertheless, a change here may be premature. This is also a topic of forum discussion.
I think you are using an adjusted SE rating as there are puzzles where the 'original' SE rating is not the same as your solver's 'SE' rating.

[StrmCkr]

Sk loop is a naked 4 digit 4 sectors 16 cells ie a naked msls thats how they are the same.

Where msls tags both naked and hidden version : the hidden set needs post move basics for clean up then it also matchs eliminations of the sk loop.

[yzfwsf]

I am unsure of the boundaries between the levels of difficulty. I think your intention was to use equivalents to the default boundaries in HoDoKu but now using the original SE ratings. What is the boundariy between Extreme and Insane?

Dim Shared As Long difLevel(0 To 6) = {0, 15, 28, 60, 80, 93, 120}
Use SKFR ,the grading boundary is as above.

Here are a few more suggestions and queries:

May I suggest adding the SK Loop? It is true that the SK Loop can overlap considerably with MSLS. pjb has noted that there are puzzles where the two steps can give different exclusions though. The logic is quite different though.

When I have time, I will implement it. In fact, I wrote SK_Loop in the previous version of VB6.

With your implementation of JE, there are the interesting cases where additional cells can be added to the "S" cells to confirm the JE pattern. Currently the solver says: ' "S" Cells Need Include xxxx True Base Cands xxxx'. I think this needs rewording. The additional cells are not "S" cells and there are no known True Base Candidates [I am unsure what is intended here?]
Since there are Naked Quads and Hidden Quads, can there be Almost Locked Quads? Maybe such an ALQ step is getting too close to Sue de Coq or ALS-XZ? The logic would be different though.
One receives the Avoidable Rectangle in solution paths. It should listed in 'Techniques'. Recent forum discussion on the AR may be relevant. [UR 1.1]. Alas, I understand very little of that discussion.

It is not complicated for a computer to detect a base candidate as follows. If it appears in the base cell, and it is assumed that the target cells does not have it, and then enumerate the positions in the cross line. If a house will miss the number in all the above situations, it is proved that as long as the number appears in the target cell at the base cell at the same time.

[yzfwsf]

In this update, due to the structural adjustment, it is estimated that bugs are inevitable. SK loop technique is added to provide program option configuration. The program can save and load configuration items. In Edit->preferences->Techniques, you can drag and drop items to Change the calling order of solving techniques.
Add filters for STE / STTE steps when searching for all possible steps.
Speed up about 100% duo to "-O3".
Please download the two files at the same time and put them in the same directory, remove ". 7z" at the end of each file name, and then decompress them.

Release 609:

YZF_SUDOKU609.7z.001.7z

(255 KiB) Downloaded 195 times

YZF_SUDOKU609.7z.002.7z

(253.48 KiB) Downloaded 198 times

[yzfwsf]

@ghfick
Because the gcc-o3 compiled version is slow to run in WINE, I'll provide a version compiled with FBC, but it's nearly 100% slower than the GCC version.
Please download the two files at the same time and put them in the same directory, remove ". 7z" at the end of each file name, and then decompress them.
release 609 fbc version for WINE

YZF_SUDOKU609-FBC.7z.001.7z

(255 KiB) Downloaded 175 times

YZF_SUDOKU609-FBC.7z.002.7z

(220.73 KiB) Downloaded 173 times

[ghfick]

In 609 [both versions], may I suggest that the default order in 'Preferences' end with:

JE
SK Loop
MSLS
Senior Exocet
Complex {Fish}
Dynamic Chain
Brute Force

Please change the ratings too. Maybe to JE=500, SK Loop=510, MSLS=520, Senior Exocet=530, Complex {Fish} >530, Dynamic Chain=1000, Brute Force=10000

I am unable to change the defaults. If I attempt to alter the order, I get messages like 'teq over'. I must then rebuild YZF_Sudoku from the 7z files.

In 'Preferences', change Ufair to Unfair and Als to ALS

As YZF_Sudoku 609 launches, one gets : Config FILE data error, load default settings

[yzfwsf]

In 609 [both versions], may I suggest that the default order in 'Preferences' end with:

JE
SK Loop
MSLS
Senior Exocet
Complex {Fish}
Dynamic Chain
Brute Force

Please change the ratings too. Maybe to JE=500, SK Loop=510, MSLS=520, Senior Exocet=530, Complex {Fish} >530, Dynamic Chain=1000, Brute Force=10000

I am unable to change the defaults. If I attempt to alter the order, I get messages like 'teq over'. I must then rebuild YZF_Sudoku from the 7z files.

In 'Preferences', change Ufair to Unfair and Als to ALS

As YZF_Sudoku 609 launches, one gets : Config FILE data error, load default settings

Maybe it failed to write the configuration file in your Windows, please modify the configuration file manually, because in my windows xp system, the program runs normally, and I can't know where there is a problem. I provide a configuration file that you can manually configure to meet your requirements.
However, when the dynamic chain is adjusted to the end, the solving time of the extreme puzzle will increase obviously.

YZF_Sudoku.ini

(1.59 KiB) Downloaded 155 times

[ghfick]

no luck with the .ini file
I am going to search for a machine with Windows XP on it.
I am unsure why 609 will not run properly with Wine. I am going to try various possible fixes. I guess I have the time right now [sigh]
Even if the solver runs appreciably slower with Dynamic Chains at the bottom of the list, I would still argue that JE, SK Loop and MSLS are steps that humans can [in principle] use. The vary elaborate Dynamic Chains are not usable by humans. I see them as a Last Resort and just before Brute Force.

[yzfwsf]

no luck with the .ini file
I am going to search for a machine with Windows XP on it.
I am unsure why 609 will not run properly with Wine. I am going to try various possible fixes. I guess I have the time right now [sigh]
Even if the solver runs appreciably slower with Dynamic Chains at the bottom of the list, I would still argue that JE, SK Loop and MSLS are steps that humans can [in principle] use. The vary elaborate Dynamic Chains are not usable by humans. I see them as a Last Resort and just before Brute Force.

I updated the way to save the configuration. Instead of using the API, I wrote the configuration to a file directly，try it please！

Please download the two files at the same time and put them in the same directory, remove ". 7z" at the end of each file name, and then decompress them.

YZF_SUDOKU609V2.7z.001.7z

(256 KiB) Downloaded 166 times

YZF_SUDOKU609V2.7z.002.7z

(254.89 KiB) Downloaded 175 times

[ghfick]

Looks promising. I manually edited the .ini file so that the end of 'Sort of Techniques' is :

40=40
41=41
42=42
43=47
44=48
45=49
46=50
47=44
48=45
49=43
50=46
51=51

I have just begun testing but 609 appears to have made the changes. The only artifact I have noticed so far is that the Dynamic Chains are now white instead of red in both the Path and Steps tabs.

Also, if I click on an alternate in Steps, 609 appears to hang. The very bottom panel flickers. Then 609 responds after a delay.

For Insane puzzles, you are right that Auto Solve is noticeably slower with the change in order but this is a very minor matter and not an inconvenience.

[yzfwsf]

I have just begun testing but 609 appears to have made the changes. The only artifact I have noticed so far is that the Dynamic Chains are now white instead of red in both the Path and Steps tabs.

You can manually configure other options as well.Section [SortOfTechniques],[RatingsOfTechniques],[LevelOfTechniques],[ColorOfMarker]...
Where [LevelOfTechniques] is to change the backcolor of solution path panel.
Do you try 609V2? That version use direct IO to save configuration.

Also, if I click on an alternate in Steps, 609 appears to hang. The very bottom panel flickers. Then 609 responds after a delay.

If you dbclick item in "All possible steps", the solver will search another solution path,I don't know if that's what you're talking about.

[ghfick]

I have been exploring the .ini file with some success. I am able to edit the file but sometimes 609 appears to change the file as well. All seems well.
Your software has many features. Very impressive and very interesting. I now see that when ones clicks on an alternate step, the solution path is reset using the change in the exclusions determined by the step. Also very interesting and this reset appears to explain the delay in response.
I am now back to testing 609 and I should have some further suggestions for you in a while. I have noticed a step called 'ERI Pair'. I am not familiar with this step. The logic is using 'Empty Rectangle like' logic. So interesting. Reference, please.