The New Sudoku Players' Forum

[mith]

The 36 million is the 4+ million T&E(3) minimals + whatever was in my local copy of ph; there are relatively few trivalue oddagon puzzles in there.

The seed for the T&E(3) database was about 900 puzzles (I have them in a spreadsheet somewhere), 11.3+ SER, all found by me neighborhood searching starting from the ph2010 database (plus a relatively small number of other puzzles - I had been investigating 19c and had some of the 9.X puzzles with that diagonal pattern that was explored previously; anyway, the other puzzles likely didn't contribute much if anything to the search, and the T&E(3) puzzles almost certainly come from the trivalue oddagon puzzles already existing in ph2010).

Anyway, I still have all of that, I'll figure out some sensible way to make it available.

I ran a quick check against the threshold, since my threshold for keeping puzzles was lower (and when I started, I was also looking for high q2 and q1 puzzles so the focus wasn't entirely on skfr as a filter). Here's a summary of what's in there:

Code: Select all

clues SER>=10.5    skfr    total
19            3       0        3
20         1527     145     1672
21        17692   21092    38784
22       119409  107384   226793
23       441375  353999   795374
24       778355  512746  1291101
25       597191  626208  1223399
26       901382  944165  1845547
27       415914 1183307  1599221
28       188980 1058993  1247973
29       101693  882608   984301
30        90912  607298   698210
31        88350  311076   399426
32        47914   74615   122529
33         6112   12248    18360
34          438     684     1122
35          330       1      331
36            4       0        4
total                   10494150

(the "skfr" column is puzzles not meeting the SER threshold but with skfr>=10.5; probably many these are unrated by serate and may have skfr significantly higher than the threshold)

[mith]

Also, what tool do you use for your canonical form? I had been using gsf's minlex (which is not true minlex, as discussed elsewhere), so that's what's in my databases, but I know you do a maxlex form instead.

[champagne]

The 36 million i..
Anyway, I still have all of that, I'll figure out some sensible way to make it available.

on my side, any text file of the puzzles is a good start
and 36 millions is less than some of my files out of the vicinity steps

[champagne]

Also, what tool do you use for your canonical form? I had been using gsf's minlex (which is not true minlex, as discussed elsewhere), so that's what's in my databases, but I know you do a maxlex form instead.

for several reasons, I changed my canonical morph.

as discussed in this thread
http://forum.enjoysudoku.com/high-density-files-for-solution-grids-and-18-clues-puzzles-t42669.html
I was looking for shorter morphs of the puzzles what was needed for huge darabases
I wanted something easy to use with the sort and other coreutil tools

On the other way, the solution grid appeared more and more as a key property of the puzzle.
I also had meantime implemented my virtual catalog giving an easy and quick {solution grid <=> rank}

All this together pushed me to the canonical morph described in this thread

Code: Select all

http://forum.enjoysudoku.com/proposal-for-a-canonical-puzzle-t45209.html

Depending on the use, a puzzle can now be shown as

The ED puzzle
a canonical name 25 bytes
a canonical name 19 bytes. This surely for local huge databases. My "known puzzles" database is now close to 300 million puzzles.

I use my own code but as explained in the last link, I have to build a DLL to replace my github release. This can be done in one or 2 days if it becomes urgent.

[champagne]

In the former PH database, without human error, a sudoku entering the base could not disappear.

Here, the situation is slightly different.

Once a solution grid has a hit, it will stay in the base, but within the solution grid, a new expanded puzzle can be subset of previous one.

Then the newly expanded sudoku will replace the old one
I have nearly finished my work and controls on the first addition to the dy_ph_th_2510 base.

Addition starts with 339 971 new minimal giving 42 933 new expanded.

At the end, 3948 puzzles of the old base are replaced by new ones.
If anybody is interested, I’ll give examples of such replacements

[champagne]

The update of the TH base is on my google drive

dy_ph_th_2511

As this was my first run of the update process, I only took new puzzles attached to an exixting solution grid of thr TH base, and yes, the count of solution grids is stable here.

Next step will be with all new puzzles, but I want before to see what is in solution grids having both TH and not TH puzzles.

[champagne]

first investigations in solution grids having currently puzzles in 'TH' and in 'noTH' show interesting points.

Part of the solution grids have in the 'noTH' side some degenerated TH as here first seen as 'TH' second as 'noTH'

Code: Select all

.2.4..7..........3.8...1..4....93.47.3714....94.5.731.39..14.75..197.43....3.5...   200719884
.2....7..........3.8.....6.....93.47.3714....94.5.731.39..14.75..197.43....3.5..1   200719884


and some with quite different cases

Code: Select all

...45............2...1.2.....8...97.5.7....2669..2.8.5.629.8.578.52..69.97..6.2.8   607395124   O0WeWO5UgkkPJ5   109;134;2;0
...4.67..........2...1.2.....8..597.5.7....2669..2.8.5.629.8.578.52..69.97....2.8   607395124   e1WeWS5UgkkP35   110;134;2;0

.....6.8...6.8.13..8.13...4.3........4...132....32...5.6...845.8...7....9..5.....   607395124   WYQQK0Y3ZYB290   26   105;0;0;0
.....6.8.4.6.8.13..8.13.....3........4...132....32...5.6...845.8...7....9..5.....   607395124   WgQQG0Y3ZYB290   26   106;0;0;0


many of theese puzzles could be of interest for players, but analysing all 454 cases will take time.

I'll post in the puzzle section some of solution grids cases

[champagne]

slow process partly due to other tasks, but with a new recognition of the potential TH threat, I reduced significantly the count, mainly after recognition of degenerated TH threat.

The change is mainly in the 'assumed non TH' file where from 112296 puzzles I have now 9728 in 151 solution grids.

Likely, most of them have a 'non TH' threat as in the former example.

I have to see what I can do to avoid analyzis of each of theese 9728 puzzles (or 151 solution grids)

BTW, It is amazing to see such a small number of solution grids common to both files.