SE:
Hidden Text: Show
SJE:
Hidden Text: Show
- Code: Select all
...4.9...31.....57....4.............8...5...6.............2....79.....23...3.5... 111111123144552223145522663445226633455266337456668397458888397488999997889977777
difference in rating between SE and SJE
18 posts
• Page 1 of 2 • 1, 2
difference in rating between SE and SJE
by urhegyi » Sun Oct 24, 2021 2:28 pm
I found already differences in rating between both programs. Now I found an example to reproduce this which has to do with the way they handle generalized intersections.
SE:
SJE:
SE:
SJE:
- urhegyi
- Posts: 748
- Joined: 13 April 2020
Re: difference in rating between SE and SJE
by Mathimagics » Mon Oct 25, 2021 2:47 pm
I think that "2.9 Generalized Intersections" is a Jigsaw-only technique? Perhaps it corresponds to what we commonly call "LoL" eliminations.
That would explain its absence from the SE(Jigsaw hack) ratings ...
As 1to9only said, the SE/SJE ratings will be different, for precisely this reason.
That would explain its absence from the SE(Jigsaw hack) ratings ...
As 1to9only said, the SE/SJE ratings will be different, for precisely this reason.
-
Mathimagics - 2017 Supporter
- Posts: 1926
- Joined: 27 May 2015
- Location: Canberra
Re: difference in rating between SE and SJE
by 1to9only » Mon Oct 25, 2021 3:43 pm
In SE when Latin Square is selected there are no 3x3 blocks only rows and columns so General Intersection is not needed!
This also applies when other variants are enabled!!
Will need to look into this further!!!
This also applies when other variants are enabled!!
Will need to look into this further!!!
-
1to9only - Posts: 4177
- Joined: 04 April 2018
Re: difference in rating between SE and SJE
by urhegyi » Mon Oct 25, 2021 7:29 pm
Another example to reproduce:
SE rating 6.6
SJE rating 2.9
- Code: Select all
.74........9.....5....6..14...........3...6...........24..3....7.....9........19. 111111123144552223145522663445226633455266337456668397458888397488999997889977777
SE rating 6.6
- Code: Select all
Analysis results
Difficulty rating: 6,6
This Sudoku can be solved using the following logical methods:
57 x Hidden Single (1,0-1,5)
4 x Direct Claiming (1,9)
1 x Direct Hidden Pair (2,0)
3 x Naked Single (2,3)
17 x Claiming (2,8)
1 x Naked Pair (3,0)
1 x X-Wing (3,2)
1 x XY-Wing (4,2)
3 x Generalized X-Wing (6,5)
4 x Forcing X-Chain (6,6)
SJE rating 2.9
- Code: Select all
Analysis results
Difficulty rating: 2,9
This Sudoku Jigsaw can be solved using the following logical methods:
54 x Hidden Single (1,0-1,5)
5 x Direct Pointing (1,7)
1 x Direct Claiming (1,9)
2 x Direct Hidden Pair (2,0)
3 x Naked Single (2,3)
10 x Pointing (2,6)
9 x Claiming (2,8)
8 x Generalized Intersection (2,9)
- urhegyi
- Posts: 748
- Joined: 13 April 2020
Re: difference in rating between SE and SJE
by Hajime » Mon Oct 25, 2021 8:48 pm
Can be solved with hidden/naked pairs , no triples and quads needed.
AND if you allow generalized intersections. No other methods needed.
Boxes and Jigsaw shapes are so different that generalized intersections are possible.
A SER rating of 3.0 is the outcome because I don't differ between the Naked Pair and the Direct Pair.
But I think generalized intersections are under estimated.
Can you see that (6)r2c2J5 => (-6)r6c2 (where J5 is Jigsaw piece 5) ? I surely can not.
SJE is correct.
Summary of logging:
Possible solution path:
AND if you allow generalized intersections. No other methods needed.
Boxes and Jigsaw shapes are so different that generalized intersections are possible.
A SER rating of 3.0 is the outcome because I don't differ between the Naked Pair and the Direct Pair.
But I think generalized intersections are under estimated.
Can you see that (6)r2c2J5 => (-6)r6c2 (where J5 is Jigsaw piece 5) ? I surely can not.
SJE is correct.
Summary of logging:
- Code: Select all
Eliminated candidates per Method and per Sudoku
Method \ Sudoku | SER | 1
|-------|------
Not counted elims | 0 | 108
Naked Singles | 0.1 | 34
Hidden Singles | 0.2 | 102
Naked Pair [2] | 3 | 5
Pointing/Claiming | 2.8 | 66
Locked Singles [3] | 2.9 | 30
|-------|------
Eliminated Cand's | 345 | 345
Sum(SER * Cand's) | 310.6 | 310.6
Initial Candidates : 345
Maximum SER rating : 3 <- Approach
Labour rating : 310.6 <- Experimental rating
Time needed : 00:00:02.169
SiSeSuSo Solver and Generator (version 2021-10)
Possible solution path:
Hidden Text: Show
-
Hajime - Posts: 1375
- Joined: 20 April 2018
- Location: Fryslân
Re: difference in rating between SE and SJE
by 1to9only » Tue Oct 26, 2021 7:19 am
Another quick fix (2021.10.26). This should fix the General Intersection problem.
There are other improvements as well, but the changes have not been tested much!
There are other improvements as well, but the changes have not been tested much!
-
1to9only - Posts: 4177
- Joined: 04 April 2018
Re: difference in rating between SE and SJE
by urhegyi » Tue Oct 26, 2021 4:44 pm
Did a few tests with the 26-10-21 new pre release version.
shows now this(SE 6.6):
shows now this(SE 2.9):
Seems to be correct now, but the major improvement is that you don't have to edit the json file anymore. Pasting a new custom layout and selecting some variants can now be done from the menu.
- Code: Select all
...4.9...31.....57....4.............8...5...6.............2....79.....23...3.5... 111111123144552223145522663445226633455266337456668397458888397488999997889977777
shows now this(SE 6.6):
Hidden Text: Show
.png)
- Code: Select all
.74........9.....5....6..14...........3...6...........24..3....7.....9........19. 111111123144552223145522663445226633455266337456668397458888397488999997889977777
shows now this(SE 2.9):
Hidden Text: Show
.png)
Seems to be correct now, but the major improvement is that you don't have to edit the json file anymore. Pasting a new custom layout and selecting some variants can now be done from the menu.
- urhegyi
- Posts: 748
- Joined: 13 April 2020
Re: difference in rating between SE and SJE
by Mathimagics » Wed Oct 27, 2021 6:55 am
1to9only wrote:Another quick fix (2021.10.26). This should fix the General Intersection problem.
There are other improvements as well, but the changes have not been tested much!
Looks good to me!
-
Mathimagics - 2017 Supporter
- Posts: 1926
- Joined: 27 May 2015
- Location: Canberra
Re: difference in rating between SE and SJE
by Mathimagics » Wed Oct 27, 2021 8:06 am
I notice that "serate" as command-line tool doesn't seem to read the json file, unfortunately, but simply relies on options to specify variants. Something for the future, perhaps?
For custom layouts, only the GUI works for rating/analysis.
For custom layouts, only the GUI works for rating/analysis.
-
Mathimagics - 2017 Supporter
- Posts: 1926
- Joined: 27 May 2015
- Location: Canberra
Re: difference in rating between SE and SJE
by 1to9only » Wed Oct 27, 2021 6:17 pm
Mathimagics wrote:I notice that "serate" as command-line tool doesn't seem to read the json file, unfortunately, but simply relies on options to specify variants. Something for the future, perhaps?
The json file is always read. For command-lines (e.g. serate), all variant settings are reset (ignored!), so the dafault serate (with no options specified) is to rate a vanilla sudoku.
So serate behavior is the same in all Explainers.
For anything else, options describe the sudokus being rated.
- Code: Select all
-L LatinSquare
-X Diagonals
-D DisjointGroups
-W Windoku
-A Asterisk
-C CenterDot
-G Girandola
-H Halloween
-P PerCent
-S S-doku
-U Custom
-V OddEven
Also command-lines options are not saved to the json file. So the GUI restarts in its last state.
Mathimagics wrote:For custom layouts, only the GUI works for rating/analysis.
For cUstom and oddeVen sudokus, the json file is read, so the layout is loaded.
Specifying the -U or -V option uses the last saved layout, so command-line serate should work for these as well.
-
1to9only - Posts: 4177
- Joined: 04 April 2018
Re: difference in rating between SE and SJE
by Mathimagics » Thu Oct 28, 2021 2:10 am
1to9only wrote:For cUstom and oddeVen sudokus, the json file is read, so the layout is loaded.
Specifying the -U or -V option uses the last saved layout, so command-line serate should work for these as well.
Ok, got it! I added -U to my serate run and it then worked fine. (I was actually testing JSB's in this case, but for JS's I would need to also add the -L option, right?)
Again, well done!
It is nice to have 9x9 Jigsaw support accepted back into the mainstream ... even if smuggled in via the rear entrance.
Cheers
MM
-
Mathimagics - 2017 Supporter
- Posts: 1926
- Joined: 27 May 2015
- Location: Canberra
Re: difference in rating between SE and SJE
by 1to9only » Sat Oct 30, 2021 2:25 pm
1to9only wrote:Another quick fix (2021.10.26). This should fix the General Intersection problem.
I've checked a few of my other Explainers - so far, I've only found Sukaku6x6Explainer needs the fix applied.
-
1to9only - Posts: 4177
- Joined: 04 April 2018
Re: difference in rating between SE and SJE
by urhegyi » Sun Oct 31, 2021 7:11 pm
1to9only wrote:1to9only wrote:Another quick fix (2021.10.26). This should fix the General Intersection problem.
I've checked a few of my other Explainers - so far, I've only found Sukaku6x6Explainer needs the fix applied.
Can you explain the difference in solving approach between SE and SJE based on this example?
- Code: Select all
..9...5....1...3..78.....92...........................63.....49..4...2....6...4.. 111111123456616123456666223456622223455577323445777333445577778499998888999998888
SE:
Hidden Text: Show
.png)
SJE:
Hidden Text: Show
.png)
- urhegyi
- Posts: 748
- Joined: 13 April 2020
Re: difference in rating between SE and SJE
by 1to9only » Sun Oct 31, 2021 10:18 pm
I have not checked the example given.
If you check the solution paths, they will be different. This is due to the solving order of Singles and Subsets and Intersections.
In SJE:
- Blocks (3x3) (if JSB), then Jigsaws (Irregular Regions) are checked - rating ED=1.2.
- Then Columns, Rows, and Variants (X) - rating ED=1.5
In SE:
- Blocks (3x3) (if not Latin Square) are checked - rating ED=1.2.
- Then Columns, Rows, and Variants (X,W,DG,etc. AND Jigsaws (Extra Regions)) - rating ED=1.5.
The pencilmarks are different enough that the grid is solved differently by SE/SJE. Otherwise the code for SE and SJE are very similar.
Edited: see colored text.
If you check the solution paths, they will be different. This is due to the solving order of Singles and Subsets and Intersections.
In SJE:
- Blocks (3x3) (if JSB), then Jigsaws (Irregular Regions) are checked - rating ED=1.2.
- Then Columns, Rows, and Variants (X) - rating ED=1.5
In SE:
- Blocks (3x3) (if not Latin Square) are checked - rating ED=1.2.
- Then Columns, Rows, and Variants (X,W,DG,etc. AND Jigsaws (Extra Regions)) - rating ED=1.5.
The pencilmarks are different enough that the grid is solved differently by SE/SJE. Otherwise the code for SE and SJE are very similar.
Edited: see colored text.
Last edited by 1to9only on Tue Nov 02, 2021 3:21 pm, edited 1 time in total.
-
1to9only - Posts: 4177
- Joined: 04 April 2018
Re: difference in rating between SE and SJE
by urhegyi » Mon Nov 01, 2021 4:01 pm
1to9only wrote:I have not checked the example given.
If you check the solution paths, they will be different. This is due to the solving order of Singles and Subsets.
In SJE:
- Blocks (3x3) (if JSB), then Jigsaws (Irregular Regions) are checked - rating ED=1.2.
- Then Columns, Rows, and Variants (X) - rating ED=1.5
In SE:
- Blocks (3x3) (if not Latin Square) are checked - rating ED=1.2.
- Then Columns, Rows, and Variants (X,W,DG,etc. AND Jigsaws (Extra Regions)) - rating ED=1.5.
The pencilmarks are different enough that the grid is solved differently by SE/SJE. Otherwise the code for SE and SJE are very similar.
My question was how it comes that SE uses 3 times claiming technique(SE 2.8) and SJE 8 times pointing(SE 2.6) on the same puzzle.
- urhegyi
- Posts: 748
- Joined: 13 April 2020
18 posts
• Page 1 of 2 • 1, 2
