Non degenerated tridagon puzzles direct search

Everything about Sudoku that doesn't fit in one of the other sections

Re: solution grid 5472482993

Postby coloin » Fri Jan 10, 2025 12:44 pm

your solution grid program is working well i see...and the puzzle identifier is also good !!
I will see if I can get to grips with the it...

however , there are advantages in this scenario of using the maxlexsoutiongrid versions of the puzzle.
All 112 of those initial 11.6/11.7 puzzles come from this expand
Code: Select all
12..5678..5718.6..6.87.2....6.57....5...68..7...2.1..6...82.961.16.....5...61.4..

The puzzles that you found [same solution grid] can be traced to these expands
Code: Select all
12..56.8..5718...26.87.2.1..6.57...85...68...87.2.1......82.....16...8...8.61.4.3 ED=10.5/10.5/2.6

12..5678..57...6.26.87.251..6.57........68..7.7.2.1.....5.2.9.1.1.............47. ED=10.4/10.4/2.6

12..5678..5718.6.26.87.2.1....57........68..7...2.1..6..5.2.9.1.1..........61.4.. ED=10.5/10.5/10.2
12..5678..5718.6.26.87.2.1....57........68..7...2.1..6..5...9.1.1..........61.4.. ED=10.9/10.9/10.3

champagne wrote:BTW, with the usual technics, it seems to me difficult to get out of the “tridagon square” of the seed.

Yes its difficult to break out of the 4 boxes with the tridagon ... twin puzzles achieve this ive found though...
coloin
 
Posts: 2534
Joined: 05 May 2005
Location: Devon

Re: solution grid 5472482993

Postby champagne » Fri Jan 10, 2025 3:58 pm

coloin wrote:
however , there are advantages in this scenario of using the maxlexsoutiongrid versions of the puzzle.
All 112 of those initial 11.6/11.7 puzzles come from this expand
Code: Select all
12..5678..5718.6..6.87.2....6.57....5...68..7...2.1..6...82.961.16.....5...61.4..

The puzzles that you found [same solution grid] can be traced to these expands
Code: Select all
12..56.8..5718...26.87.2.1..6.57...85...68...87.2.1......82.....16...8...8.61.4.3 ED=10.5/10.5/2.6
12..5678..57...6.26.87.251..6.57........68..7.7.2.1.....5.2.9.1.1.............47. ED=10.4/10.4/2.6

12..5678..5718.6.26.87.2.1....57........68..7...2.1..6..5.2.9.1.1..........61.4.. ED=10.5/10.5/10.2
12..5678..5718.6.26.87.2.1....57........68..7...2.1..6..5...9.1.1..........61.4.. ED=10.9/10.9/10.3


I did not work on this, but I agree that using maxlexsoutiongrid should make it simpler

If we order the one plus four of your grids in this way

Code: Select all
12..5678..5718.6..6.87.2....6.57....5...68..7...2.1..6...82.961.16.....5...61.4..  36 clues
 
12..5678..5718.6.26.87.2.1....57........68..7...2.1..6..5.2.9.1.1..........61.4..  33 clues
12..5678..5718.6.26.87.2.1....57........68..7...2.1..6..5...9.1.1..........61.4..  32 clues

12..56.8..5718...26.87.2.1..6.57...85...68...87.2.1......82.....16...8...8.61.4.3 36 clues
12..5678..57...6.26.87.251..6.57........68..7.7.2.1.....5.2.9.1.1.............47.   32clues


it seems to me that a vicinity applied here would give your grid starting from the four seeds. But surely several steps needed.
I am surprised to see the second and third seeds differing only by one clue!!


coloin wrote:
champagne wrote:BTW, with the usual technics, it seems to me difficult to get out of the “tridagon square” of the seed.

Yes its difficult to break out of the 4 boxes with the tridagon ... twin puzzles achieve this ive found though...


If an efficient way to use the seeds appears, a good strategy would be to search first new squares, but I am not sure that it is easy to recognize the same square in different solution grids in min lexical morph.
Anyway, seeds with lower ratings must be considered. I'll post ratings 10.5 t 11.x as soon as the test is over.
champagne
2017 Supporter
 
Posts: 7523
Joined: 02 August 2007
Location: France Brittany

Re: Non degenerated tridagon puzzles direct search

Postby champagne » Sat Jan 11, 2025 4:19 pm

Hi coloin,

the last discussion on a the solution grid 5472482993 could lead to a significant improvement.

On my side, the bottleneck remains the filter to select the grids of interest in the millions grids in the output of the tridagon finder.
The T&E(1) filter kills more than 99% of the grids, but the rating of the rest is still too long and you have shown that the result is not directly used by you.

I made a test on the "quick solver" that I started when I wanted to have the T&E(1) filter. Expanding a puzzle on rules can be very fast and as you noticed, delivers many duplicates, saving time in the final rating.

I intend to introduce this filter after the T&E(1) filter. It's too late to apply it to the current test, most of the results have already been thrown to the bin, so, I'll redo a test in another tiny area later;

I can push the solving path to different points. From your experience (with no "uniqueness rule"), where is the right cutoff?? For sure what can be coded in a simple and fast way will be considered, but I see as simple in sudoku Explainer rules for example

in block all pointing/claiming eliminations
In block all fishes xwing to turbots and more
in block all unit locked sets (pairs, hidden pairs...)
In block all Ychains (includes XYwing)
....
So far, I have drafted and tested the 2 first rules and I am working on the third one.

I'll do the test in the next days with the rules ready on the 10.5-11.x ratings skfr .
champagne
2017 Supporter
 
Posts: 7523
Joined: 02 August 2007
Location: France Brittany

scab band1 range 320_349

Postby champagne » Sun Jan 12, 2025 4:55 pm

Short summary of the results for al targets in solution grids with a band 1 in the range 320-349 (index)

I got 13907 puzzles with a skfr ER rating in the range 10.5-11.2 (nothing over 11.2)
I did the expand clean using my "quick solver" code and got

3914 different expanded grids
1000 different solution grids.

Knowing that I have seen only one solution grid in mith's file in this tiny area, we can assume 999 new solution grids.

This is still too big for a post, but easy to send by mail to anybody interested in getting the entire file.
I selected as example some results after expansion and re rating.

Code: Select all
5472052039;nnDT3992Kpe5U4;8...5.....4..9...5.5.3........13..57...5.96.3..5.6791.51..73.967..6.5..16..91.57.;110

5472055810;Yhe36A844MTp53;.9..4....3....1.....1.........72..6826.4.81.7.7..1642.....62.74...1.48.6..687.2..;110

5472056429;H3M0eLsO3qeBh2;78...5.1..3.9.8.............5.89..21...4.29.5....5148.2...89.54...1.4..2.4.52.19.;105
5472056429;A9A0iHGPRqe3h0;7...4.2..6....8.4...4.......5..9..211....2...92..5148.....89.54...1.4..2...52.19.;105

5472059357;eBh3cTo5p3ZY13;34.2.8.....6.1..4.9..7.4......48..26..46.17.8.6..7241.4...26.81...8.76.468.14....;105
5472059357;f3hXETm5p5G003;3.5.......7..........7.4......48..262..6.17.8....7241.....26.811..8.76.46..14.2..;106

5472061482;8Y04ani5UPDUj6;.....8.4...7........9.4....3..48..565..3.28.48.4.5632...5.64.8....8.54.2..823.56.;109

5472061896;rmPR0G19AhgcU1;.6.......9....7...2....8..13..87..56..61.53.7.5..6318.....86.73...3.16.563.75..1.;108
5472061896;rmPR1C39AhgcU1;1.....7..9...17...2....8..13..87..56..61.53.7.5..6318.5...86.73...3.16.563.75..1.;105

5472063425;o6h3hTOck1Em90;3...19...7...56...8.....5....816..252.69.58.1....8269..2..98.5668.5.1......62.18.;111

5472066663;gG4G0m8bPpRpe0;.2....6.4.5....9.........3.87.36..49......8.334..9876...6.43..8...8.9..6...67....;107
5472066663;gG4G0o8bPpRpe0;.2....6.4.5....9.........3.87.36..49......8.334..9876...6.43..8...8.9..6...67..9.;107
5472066663;gG4G0o8rPpRpe0;.2....6.4.5....9.........3.87.36..49......8.334..9876..96.43..8...8.9..6...67..9.;105
5472066663;AK0H0o85PpRpe0;.2.9....4.5.......6......3.87.36..49......8.334..9876.....43..8...8.9..6...67..9.;111
5472066663;AK0H0o8bPpRpg0;.2.9....4.5.......6......3.87.36..49.6....8.334..9876...6.43..8...8.9..6...67..9.;110
here all results for this solution grid

5472067317;8A9GO6UfLbS3s1;.8......46....4..1.......3.9..51..42.2.4.93.5....3219.....95.13..52.34..3..14.52.;111

5472074297;p2he3CImIoTpi3;...7....1.3..1.....91.6.2...1.57..68..51.87.287..2651.....81.26...2.71.512.65.8..;109
5472074297;p2heZ4IGIm5pi3;..87......3..1.....9..6.2.....57..68..51.87.2....2651.....81.26...2.71.512.65.8..;109

5472074446;2960QfLsO56Sc5;....7..2..89.....4.....1...5..73..42.4.1.25.7....4513...4.27......5.42.3..531.47.;111

5472077643;g3hW6GqK6QMTn5;...4.......5.6.14..3....86..4.27..16.1.6.47.87...8142.....26.81.2.8.76.4...14.2..;111

5472095763;1IhfZCGGWs5pe3;..41....8....7.....3.....1..6572..81...8.56.2....1675..5..81.266..2.71.51.....8..;111

5472242347;i82TfmIY0jUMU0;.97.42....5.6....4....5.......82..56.6.4.53.2.25.6348.5...86.23.....4..8....3.64.;112

5472682358;GC2WQq5peJsH50;.5....9...8.4.......7.....2...71..94..12.47.3....9312.3...41..9.4.3.72.1...92....;110
5472682358;GC2WQs5peJsH50;.5....9...8.4.......7.....2...71..94.912.47.3....9312.3...41..9.4.3.72.1...92....;112

champagne
2017 Supporter
 
Posts: 7523
Joined: 02 August 2007
Location: France Brittany

Re: Non degenerated tridagon puzzles direct search

Postby coloin » Mon Jan 13, 2025 7:48 pm

champagne wrote:I can push the solving path to different points. From your experience (with no "uniqueness rule"), where is the right cutoff?? .


The cut off I used was SE 11.0 …. Although realistically for most puzzles it was 11.5
Most BxB 6 and above puzzles are SE 11.5 and above .
This was my “ tridagon” finder … as there is a hump in the incidence after 11.2

I expanded the puzzles maximally to maintain the SE and BxB … rarely there was a disagreement .
Expanding reduces the total number to rate. It also increases the chance that a twin puzzle is found - and these were used as new seeds … ( this tended to change the 4 boxes concerned in the tridagon .

These puzzles were in T&E 2 mostly …. The T&E 3 were likely to be less than SE 11 - but now we understand that the uniqueness glitch did that.
coloin
 
Posts: 2534
Joined: 05 May 2005
Location: Devon

Re: Non degenerated tridagon puzzles direct search

Postby champagne » Wed Jan 15, 2025 7:51 am

coloin wrote:I expanded the puzzles maximally to maintain the SE and BxB … rarely there was a disagreement .

I'll keep this as guideline for the future.
I am busy for the next weeks, but anyway, I had finished what I can do on this topic.
Her are the results of my last test.

====================================
I run the test on the first solution grids of the band1 index 60 in the catalog
and found 26000 targets in the first 55625 solution grid

Then running 6 cores in parallel to find <= 26 clues
I got 81 606 139 pure tridagon grids in 39123 seconds

Around 35 seconds per target but 3.9 seconds per solution grids. (in the proof that no 16 clues exists, Gary Mc Guire team has a similar run time per grid)

_________________________

The T&E(1) filter cleared in 114 seconds 98.24% of the grids, leaving 1 437 088 grids
Then a “solver” step expanded these grids in about 480 seconds including the sort clean of redundant
__________________________
And the final set of grids to rate has been 239 097,
0.00293% of the finder output.
Doing so, the rating runtime was similar to the finder’s runtime.
And 598 grids were rated 10.5-11.2 hitting 212 solution grids .

here are the highest ratings
Code: Select all
5.6..4..97......4...1.........54..92...7.95.39.5.2347..23.97.54.5.3.29.....45....;11.1/10.4/3.4
...9....2189.2..7........9....23..69...4.63.736..9724..7..69..4...3.27.66..74.92.;11.1/10.5/10.3
.....3...7..4...5.8......1....13..25..19.23.4....4519.....91.32.2.5.49....932.54.;11.1/11.1/10.3
.....3...7..4...5.8......1....13..25..19.23.4....4519.4...91.32.2.5.49....932.54.;11.1/11.1/10.3
...9....2189.2..7........9....23..69...4.63.7.6..9724.27..69..4...3.27.66..74.92.;11.1/11.1/10.4
...9....2189.2..7........9...723..69...4.63.7.6..9724.27..69..4...3.27.66..74.92.;11.1/11.1/10.4
...9....2189.2...........9...723..69...4.67.3.6..9724.2...69..4...7.23.66..34.92.;11.1/11.1/2.6
..6...1...8....4..73.415......79..419..5.17.3....3495.....53.19...9.75.4.9.14.3..;11.1/11.1/2.6
.6..3....9........5....8....2.38..47...7.62.88...2436.....47.8....2.34.6.4.86.73.;11.1/11.1/2.6
5.6..4..97......4...1......3..54..92...7.95.39.5.2347..23.97.54.5.3.29.....45....;11.1/11.1/3.4
4.6.3...........3.7...9....2.198..63...1.28.9....6312.....29..8...3.82.6.2.61.39.;11.1/4.5/2.6
...4..9...57.8..............1.36..98.3.8.14.6..6.9431...1.48..9.9.6.31.4...91.86.;11.2/11.2/7.6

This is likely too long for a full scan, but open the door for the search of new seeds.





champagne
2017 Supporter
 
Posts: 7523
Joined: 02 August 2007
Location: France Brittany

Re: Non degenerated tridagon puzzles direct search

Postby champagne » Sat Mar 08, 2025 9:20 am

Blue posted in another thread results of high interest on solution grids having non degenerated tridagons.
The discussion can be seen from here
http://forum.enjoysudoku.com/tridagon-t45572-30.html

I continued on my side the test on solution grids having high ratings in the old file of potential hardest. The first results are described in the same thread, and I am running a test on all solution grids of mith's file (limiting the finder to 26 clues) to compare the results with the previous sample.

I expect to get the results next week.

In parallel, I am revising the tridagon finder to clean easy puzzles at generation time. This could be the version to search other "loki" families of interest
champagne
2017 Supporter
 
Posts: 7523
Joined: 02 August 2007
Location: France Brittany

Re: Non degenerated tridagon puzzles direct search

Postby champagne » Mon Mar 10, 2025 2:22 pm

Some data in line with the previous post.

I intended for long to run a test on the “loki family”, to set the search parameters to reasonable values

. getting a runtime closer to one second per target if possible,
. giving a good chance to catch another family if any.

The “loki family” in my mith’s file covers 67539 solution grids.

67539 / 5472730538 this is 0.00123% of the solution grids.

From Blue’s work, we know that 1/3 of the solution grids can give non degenerated tridagons.

Most of them are easy to solve, but we have good chances to find other families of interest.
If we want to find these families through a direct scan of the solution grids, we must process all possible starts in 1/3 of the solution grids, likely around 2.7 billion starts.

From my 67539 solution grids, I got 103782 possible starts in 54783 solution grids.
Blue has seen 47963 solution grids with the loki start.

We have at least 2 reasons to explain the 6820 deviation:

I am not sure that we have the same file,
We know that many puzzles in mith’s file don’t have he “loki tridagon” in the final solution grid.
One example is given here.

http://forum.enjoysudoku.com/tridagon-t45572-30.html

The fact that one solution grid has no loki tridagon can not exclude another start in the same grid.


I did several tests on these 103782 starts, working on chunks of 8000 starts.

The main test was to run the code with the finder looking for puzzles of size <=26.
The average per start was around
3.2 seconds of runtime
17000 puzzles to rate

Keeping in mind that the proof that no 16 exists took about 3s per solution grid, this is feasible but requires huge power.
And here, we still have to rate the puzzles, more or less doubling the runtime using filters.

In the main test, the next steps were to clean the output of the finder from puzzles having no chance to reach the high ratings expected in the loki family.

The cleaning is done on 2 parameters:

As Mith’s file is T&E(3), all puzzles T&E(1) are discarded.
If we expand puzzles on easy moves (singles + ???), the chances of getting a high rating decrease sharply when the number of clues grows. It seems that >31 clues can be ignored without great risk to missing important puzzles.
And we have some redundancy saving rating steps.

In the test, we end with 548191 puzzles to rate, a small number of these puzzles being subsets of another puzzle as here


974..........2.9.....6.9......56..12...2.18.9....9865..1..52.....58.6...6..91.58. ED=10.2/10.2/2.6
974..........2.9.....6.9......56..12...2.18.9....9865..1..52.....58.6..16..91.58. ED=9.4/9.2/2.6

This is now a small number of grids to rate per solution grid (~5). This is good if we still have high rating hits in the loki family.
And in fact, I got 595 hits >=11.5 with the following distribution (skfr ratings)

Code: Select all
11.5   78
11.6   286
11.7   222
11.8   9


Enough to think that, with such a cutoff, we should find similar families if any.
With a cutoff on puzzles with >=30 clues, the total number of hits goes down from 595 to 291 with the distribution

Code: Select all
11.5   33
11.6   139
11.7   115
11.8   4


An option to consider although the risk of missing a new family increases.

Compared to this, the test done on solution grids with puzzles having high ratings in the potential hardest file did not deliver a rating over 11.2
Next post will show a change in the finder giving better timing results
champagne
2017 Supporter
 
Posts: 7523
Joined: 02 August 2007
Location: France Brittany

Re: Non degenerated tridagon puzzles direct search

Postby blue » Tue Mar 11, 2025 8:12 am

champagne wrote:From Blue’s work, we know that 1/3 of the solution grids can give non degenerated tridagons.

A non-degenerated tridagon" with exactly one "true" guardian.

champagne wrote:If we want to find these families through a direct scan of the solution grids, we must process all possible starts in 1/3 of the solution grids, likely around 2.7 billion starts.

It was 2,968,332,766 exactly (from the other thread).

champagne wrote:From my 67539 solution grids, I got 103782 possible starts in 54783 solution grids.
Blue has seen 47963 solution grids with the loki start.

We have at least 2 reasons to explain the 6820 deviation:

I am not sure that we have the same file,
We know that many puzzles in mith’s file don’t have he “loki tridagon” in the final solution grid.
One example is given here.

http://forum.enjoysudoku.com/tridagon-t45572-30.html

The fact that one solution grid has no loki tridagon can not exclude another start in the same grid.

Maybe the last line above, is saying the same as this, but I can't be sure, so ...

    In mith's puzzles, all of the puzzles for a given solution grid, have the same 3 tridagon digits, with (the 2 or 3) clues for those digits, in the same box.
    If the positions of those digits in the solution grid, in the 4 "opposite" boxes, correspond to a "loki-style" start, then I counted the grid.
    The difference in counts, likely comes from grids with "loki starts" that aren't used by the actual puzzles.
blue
 
Posts: 1067
Joined: 11 March 2013

Re: Non degenerated tridagon puzzles direct search

Postby champagne » Wed Mar 12, 2025 8:25 am

Changes to do in the finder:

As I wrote, the process is still too long for a full scan. Based on the test done on the loki family, the following changes should have a huge global effect, likely doing better than halving the total runtime.

For each output, in the test, 2 filters are later applied:

a)Puzzles solved in T&E(1) are ignored
b)Puzzles having >31 clues after easy solving steps (singles + ??) are ignored

Filter b) can be first applied on each potential output of the current finder. This is faster than the brute force check of the uniqueness of the grid.
Doing so, less areas of the field are explored, and several UAs appearing in the cancelled checks are not used;

Filter a) is longer than the brute force check, but can be applied after.

I made a first attempt to implement this, with good results, but something is wrong (lost puzzles of interest) and I need free time to see what I did wrong.

=================================
blue wrote:
champagne wrote:From Blue’s work, we know that 1/3 of the solution grids can give non degenerated tridagons.

A non-degenerated tridagon" with exactly one "true" guardian.


I did not check one true guardian, but the process applied in very close.

My canonical morph for a start is

Code: Select all
... ... ...
... ... ...
... ... ...

... ..x x..
... .x. .x.
... x.. ..x

... x.. z..
... .x. .z.
... ..x ..z


Where "x" must contain in the star PM the three digits 123 and nothing more
Each “z” must contain the three digits 123, one of the 3 “z” can have other digits.

Compared to your wording, the only difference I see is the possibility to have more than one guardian in the “z” cell not limited to the digits 123

To get this, we can not have a 2 digits UA for the three digits 123 having no hit in box 1. As first condition to search a start, I select all triplets where each pair has only one UA;
And as you write 2 of the 3 digits must be given in box 1.
In fact, none of the cells of these 3 UAs can be assigned out of the box 1


blue wrote:In mith's puzzles, all of the puzzles for a given solution grid, have the same 3 tridagon digits, with (the 2 or 3) clues for those digits, in the same box.
If the positions of those digits in the solution grid, in the 4 "opposite" boxes, correspond to a "loki-style" start, then I counted the grid.
The difference in counts, likely comes from grids with "loki starts" that aren't used by the actual puzzles.[/list]


This seems to me fully in line with my code.
champagne
2017 Supporter
 
Posts: 7523
Joined: 02 August 2007
Location: France Brittany

Re: Non degenerated tridagon puzzles direct search

Postby blue » Thu Mar 13, 2025 12:10 am

champagne wrote:Each “z” must contain the three digits 123, one of the 3 “z” can have other digits.

Compared to your wording, the only difference I see is the possibility to have more than one guardian in the “z” cell not limited to the digits 123

The guardian that matches the solution grid, is the one true guardian.

champagne wrote:To get this, we can not have a 2 digits UA for the three digits 123 having no hit in box 1. As first condition to search a start, I select all triplets where each pair has only one UA;

You say "as a first condition". It isn't a sufficient coundition, though, right:

Code: Select all
+-------+-------+-------+
| 3 4 2 | 5 1 6 | 7 8 9 |
| 8 6 5 | 7 4 9 | 1 2 3 |
| 9 7 1 | 2 8 3 | 5 4 6 |
+-------+-------+-------+
| 6 3 8 | 4 9 1 | 2 5 7 |
| 7 1 9 | 8 2 5 | 6 3 4 |
| 5 2 4 | 3 6 7 | 8 9 1 |
+-------+-------+-------+
| 4 9 6 | 1 5 8 | 3 7 2 |
| 2 8 7 | 6 3 4 | 9 1 5 | digitits 1,2,3
| 1 5 3 | 9 7 2 | 4 6 8 |
+-------+-------+-------+

Removing clues for {1,2,3} in all but box 1, leaves a multi-soluton puzzle, but removing 1's and 2's only (or 1's and 3's, or 2's and 3's), leaves a single solution puzzle,
blue
 
Posts: 1067
Joined: 11 March 2013

Re: Non degenerated tridagon puzzles direct search

Postby coloin » Thu Mar 13, 2025 12:30 pm

Over past few weeks I have found more TE3 puzzles....
Code: Select all
typical single guardian  TE3 puzzle expanded maximally - clue count - {975454110}

+---+---+---+
|86.|..4|...|
|.95|8.7|..4|
|...|6..|..8|
+---+---+---+
|.78|94.|...|
|6.9|7.5|48.|
|54.|.68|9..|
+---+---+---+
|98.|...|..5|
|...|.76|819|
|...|.89|2..|
+---+---+---+      Guardian cell is 7@r1c3

+----------------------+----------------------+----------------------+
| 8      6      123-7G | 123    59     4      | 1357   23579  1237   |
| 123    9      5      | 8      123    7      | 136    236    4      |
| 12347  123    12347  | 6      59     123    | 1357   23579  8      |
+----------------------+----------------------+----------------------+
| 123    7      8      | 9      4      123    | 1356   2356   1236   |
| 6      123    9      | 7      123    5      | 4      8      123    |
| 5      4      123    | 123    6      8      | 9      237    1237   |
+----------------------+----------------------+----------------------+
| 9      8      123467 | 1234   123    123    | 367    3467   5      |
| 234    235    234    | 2345   7      6      | 8      1      9      |
| 1347   135    13467  | 1345   8      9      | 2      3467   367    |
+----------------------+----------------------+----------------------+

+---+---+---+
|..G|3..|.21|
|1..|.2.|3..|
|.23|..1|...|
+---+---+---+
|2..|..3|1..|
|.3.|.1.|..2|
|..1|2..|.3.|
+---+---+---+
|...|132|...|
|3.2|...|.1.|
|.1.|...|2.3|
+---+---+---+    3-template    completable in 2 clues

+---+---+---+
|867|.94|5..|
|.95|8.7|.64|
|4..|65.|798|
+---+---+---+
|.78|94.|.56|
|6.9|7.5|48.|
|54.|.68|9.7|
+---+---+---+
|984|...|675|
|.5.|476|8.9|
|7.6|589|.4.|
+---+---+---+  27 ways to add 1 clue and 18 ways to add the next [/2]

Way back dobrichev worked out the 3 templates completable in only 2 clues
dobrichev wrote:

Code: Select all
|k|  Template   | Rookery |   Min. Clues to Complete     |
| |   classes   | classes |   Rookery    |   Template    |
|--------------------------------------------------------|
|0|           1 |       1 |  0       (1) |  0        (1) |
|--------------------------------------------------------|
|1|           1 |       1 |  0       (1) |  0        (1) |
|--------------------------------------------------------|
|2|         181 |     170 |  4       (3) |  4        (3) |
| |             |         |  3      (16) |  3       (18) |
| |             |         |  2      (46) |  2       (55) |
| |             |         |  1     (105) |  1      (105) |
|--------------------------------------------------------|
|3|      259272 |   92048 |  6       (2) |  6        (5) |
| |             |         |  5      (26) |  5      (229) |
| |             |         |  4    (1128) |  4    (14779) |
| |             |         |  3   (24240) |  3   (116554) |
| |             |         |  2   (66652) |  2   (127705) |


Our tridagon puzzles all have the essential structure of the diagonal gaps in B1B2B4B5 and 2 of these 3 clues actual puzzle clues in B9

It seems that the vast majority of the TE3 puzzles have the clue distribution {xxxxxx110}
and I think that the ones which are {xxxxxx111} also have the 3-template which envolves the tridagon solvable in 2 clues. [all 3 clues are in B9]

@ blue - Maybe you did use this fact in the production of your table of tridagon grids.
Code: Select all
| 1807371808 | 2968332766 |

But if you didnt there maybe is a reduction by ~50%

Regarding the ED patterns of the tridagon clues in B1B2B4B5 I counted 12 ways to display B1B2B4B5 and the guardian cell is implied in the pattern. [Duplicates removed after correction edit.]
Code: Select all
...1.....13..2......2..3...3....1....2..3......12................................#ED
...1......3..2....1.2..3...3....1....2..3......12................................#ED
...1......3..2.....12..3...3....1....2..3......12................................#ED
.3.1......1..2......2..3...3....1....2..3......12................................#ED
..31......1..2......2..3...3....1....2..3......12................................#ED
...1......13.2......2..3...3....1....2..3......12................................#ED
       
32.1.........2......1..3...2....1....1..3......32................................#ED
3.21.........2......1..3...2....1....1..3......32................................#ED
3..1.........2.....21..3...2....1....1..3......32................................#ED
...1.....13..2......2..3...2....1....1..3......32................................#ED
...1......31.2......2..3...2....1....1..3......32................................#ED
...1......3..2....1.2..3...2....1....1..3......32................................#ED


Code: Select all
+---+---+---+
|..G|2..|1.3|
|1.2|.3.|...|
|.3.|..1|2..|
+---+---+---+
|2..|453|.1.|
|.1.|627|.3.|
|..3|189|..2|
+---+---+---+
|3.1|..2|...|
|...|.1.|32.|
|.2.|3..|..1|
+---+---+---+  641159 sols [641159 solution grids have this potential tridagon]
Last edited by coloin on Mon Mar 17, 2025 11:12 am, edited 1 time in total.
coloin
 
Posts: 2534
Joined: 05 May 2005
Location: Devon

Re: Non degenerated tridagon puzzles direct search

Postby champagne » Thu Mar 13, 2025 5:24 pm

coloin wrote:Over past few weeks I have found more TE3 puzzles....
[code]typical single guardian TE3 puzzle expanded maximally - clue count - {975454110}

+---+---+---+
|86.|..4|...|
|.95|8.7|..4|
|...|6..|..8|
+---+---+---+
|.78|94.|...|
|6.9|7.5|48.|
|54.|.68|9..|
+---+---+---+
|98.|...|..5|
|...|.76|819|
|...|.89|2..|
+---+---+---+ Guardian cell is 7@r1c3


Hi coloin,

Nice example rated only 10.4 by skfr, but with a UR step
Is your fresh file always in the vicinity of the "loki family" ??

I am short in time these days, and my priority remains to draft the changes in the finder to get a code speedy enough to start the scan of some areas of the solution grids field. I keep mith's file as basis to build and test my code, but I downloaded your file of expanded puzzles to check some points later.
champagne
2017 Supporter
 
Posts: 7523
Joined: 02 August 2007
Location: France Brittany

Re: Non degenerated tridagon puzzles direct search

Postby blue » Fri Mar 14, 2025 4:15 am

Hi Colin,

coloin wrote:Way back dobrichev worked out the 3 templates completable in only 2 clues
dobrichev wrote:

Code: Select all
|k|  Template   | Rookery |   Min. Clues to Complete     |
| |   classes   | classes |   Rookery    |   Template    |
|--------------------------------------------------------|
|0|           1 |       1 |  0       (1) |  0        (1) |
|--------------------------------------------------------|
|1|           1 |       1 |  0       (1) |  0        (1) |
|--------------------------------------------------------|
|2|         181 |     170 |  4       (3) |  4        (3) |
| |             |         |  3      (16) |  3       (18) |
| |             |         |  2      (46) |  2       (55) |
| |             |         |  1     (105) |  1      (105) |
|--------------------------------------------------------|
|3|      259272 |   92048 |  6       (2) |  6        (5) |
| |             |         |  5      (26) |  5      (229) |
| |             |         |  4    (1128) |  4    (14779) |
| |             |         |  3   (24240) |  3   (116554) |
| |             |         |  2   (66652) |  2   (127705) |


Our tridagon puzzles all have the essential structure of the diagonal gaps in B1B2B4B5 and 2 of these 3 clues actual puzzle clues in B9

It seems that the vast majority of the TE3 puzzles have the clue distribution {xxxxxx110}
and I think that the ones which are {xxxxxx111} also have the 3-template which envolves the tridagon solvable in 2 clues. [all 3 clues are in B9]

@ blue - Maybe you did use this fact in the production of your table of tridagon grids.
Code: Select all
| 1807371808 | 2968332766 |

But if you didnt there maybe is a reduction by ~50%

You are correct about the "ones which are {xxxxxx111}",
For what I did, the number of clues that might be kept in b9 (later on) didn't enter into play.
(In other words, no reduction is possible).

BTW:

coloin wrote:Regarding the ED patterns of the tridagon clues in B1B2B4B5 I counted 14 ways to display B1B2B4B5 and the guardian cell is implied in the pattern.
Code: Select all
...1.....13..2......2..3...3....1....2..3......12................................#ED
...1......3..2....1.2..3...3....1....2..3......12................................#ED
...1......3..2.....12..3...3....1....2..3......12................................#ED
.3.1......1..2......2..3...3....1....2..3......12................................#ED
..31......1..2......2..3...3....1....2..3......12................................#ED
...1......13.2......2..3...3....1....2..3......12................................#ED
       
32.1.........2......1..3...2....1....1..3......32................................#ED
3.21.........2......1..3...2....1....1..3......32................................#ED
3..1.........2.....21..3...2....1....1..3......32................................#ED
...1.....13..2......2..3...2....1....1..3......32................................#ED
...1......31.2......2..3...2....1....1..3......32................................#ED
...1......3..2....1.2..3...2....1....1..3......32................................#ED
.321.......1.2.........3...2....1....1..3......32................................#ED
.3.1.........2....12...3...2....1....1..3......32................................#ED

These aren't all ED.
#13 is isomorphic to #8.
#14 is isomorphic to #11.
blue
 
Posts: 1067
Joined: 11 March 2013

Re: Non degenerated tridagon puzzles direct search

Postby champagne » Mon Mar 17, 2025 8:46 am

Changes to do in the finder:

I started a first version of the changes to do in the finder.
As usual, good and bad news.

I limited the filter on potential solution to the following

Solve first on what can be done in the brute force with no guess (singles plus locked in row/box)
Ignore all solutions with >32 clues
Solve and find multi solutions (normal step in the previous code)
If it is still valid, check if it is T&E(1), then ignore it.


This gave very good results on my sample of 8000 starts derived from the loki family.
The process run time fell from 3.46 seconds per start to 2.43 seconds per start,
About 95% of the previous output vanished with a residual average of 61 puzzles in output (expanded puzzles)
Reduced to 27 puzzles per output cleaning redundant puzzles in the expanded status.

All this sounds good, unhappily, the rating of the 217263 puzzles took close to 9 hours, 3.93 seconds per start.
I already noticed that after expansion we had many residual grids with a potential for more reduction.
Here is an example

.................894.8.31..2.138..658..6.23.13...5182.....65......1.86.26..23....;31
.................894.8.31..2.138..658..6.23.13...5182..2..65......1.86.26..23....;32
.................894.8.31..2.138..658..6.23.136..5182.....65......1.86.26..23....;32
.................894.8.31.62.138..658..6.23.13...5182.....65......1.86.26..23....;32

These four grids came into the current process. Due to the initial set of given, we have at the end one grid subset of the three other grids.
I have now to find the best way to take advantage of such cases.

Note: As the cleaning of the output could be done in a separate step, the scan of significant areas could start.
champagne
2017 Supporter
 
Posts: 7523
Joined: 02 August 2007
Location: France Brittany

PreviousNext

Return to General