The New Sudoku Players' Forum

[Draco]

Recently generated this spiral puzzle (24 spot and sort of a Milky Way spiral shape) that my solver can't crack without my making a guess (don't do advanced fish, ALS, Nice Loops but does handle forcing nets from b/b squares).

Interested in solutions and thoughts on the layout pattern. I know there's been a huge thread on patterns and confess I have not combed it looking for this shape.

Code: Select all

5 8 . | . . . | . . .
. . . | . . . | 5 6 .
7 . . | . 9 1 | . . .
------+-------+------
2 . . | 4 . . | 7 . .
. 4 . | 2 . 5 | . 1 .
. . 6 | . . 3 | . . 2
------+-------+------
. . . | 1 4 . | . . 9
. 1 2 | . . . | . . .
. . . | . . . | . 8 4


Solver stalls here after SSTS and one forcing chain (r4c9=5 [r6c8<>5] r6c2=5 + r8c9=5 [r8c5<>5] r9c5=5 ==> r9c2<>5):

Code: Select all

5   8   134   | 367   2367  2467 | 349  3479 137
139 2   1349  | 378   378   478  | 5    6    137
7   6   34    | 5     9     1    | 2348 234  38 
--------------+------------------+---------------
2   359 13589 | 4     168   689  | 7    359  3568
389 4   3789  | 2     678   5    | 3689 1    368
189 579 6     | 789   178   3    | 489  459  2   
--------------+------------------+---------------
368 357 3578  | 1     4     67   | 236  2357 9   
4   1   2     | 36789 35678 6789 | 36   357  3567
369 379 3579  | 367   23567 267  | 1    8    4   


Coloring on 9's lead me to try the assumption that r4c6<>9; while that does not crack the puzzle to SSTS, it does pry it open far enough that some forcing chains/nets can.

Cheers...

- drac

[submacrolize]

Just go here.

Click on the gray squares that you want to see. Right click to view the whole solution

[udosuk]

Just go here.

Click on the gray squares that you want to see. Right click to view the whole solution

submarcrolize, anybody with a decent solver program can find the solution within seconds:

581624937
924738561
763591248
258416793
349275816
176983452
835147629
412869375
697352184

It's the application of logical techniques to solve it that matters!

Draco, this is not easy. Hopefully the chain experts will give you some help.

[daj95376]

Draco: While you're waiting for an informed response! Your elimination [r9c2]<>5 is also a finned X-Wing elimination.

Your PM has many network eliminations, but I didn't find any that cracked the puzzle. Here are a couple examples.

Code: Select all

6r1c5 6r4c6 7r7c6 2r9c6 -2r1c6 ~2r1c5 => [r1c5]<>6

7r7c3 7r5c5 7r6c2 5r6c8 5r7c2 5r4c3 1r4c5 6r4c6 -6r7c6 ~7r7c6 => [r7c3]<>7


[gsf]

Interested in solutions and thoughts on the layout pattern. I know there's been a huge thread on patterns and confess I have not combed it looking for this shape.

to check if a pattern is new to the patterns game
first find a low ER variant puzzle with the same pattern and then
send a private message to g.r.emlin with pattern as the subject and the 81-char grid as the message body
if the pattern has not been seen (and it meets the other patterns game criteria) it will be queued up for a future game
a reply message will tell if it was queued or why it was rejected

here is a 1.2/1.2/1.2 (ER / pearl / diamond) variant of you pattern you can try

Code: Select all

520000000000000560700091000800400600040205010006003009000140005012000000000000082

[Draco]

to check if a pattern is new to the patterns game
first find a low ER variant puzzle with the same pattern and then
send a private message to g.r.emlin with pattern as the subject and the 81-char grid as the message body
if the pattern has not been seen (and it meets the other patterns game criteria) it will be queued up for a future game
a reply message will tell if it was queued or why it was rejected

Hi Glenn,

Not even sure what a low ER variant puzzle is. I've seen some of the submissions but haven't read the rules. Must admit I am less interested in entering a puzzle than I am in finding out techniques (short of recursion, which trivially cracks the puzzle, or educated guessing which I used when I was stuck) that might crack it and if this was/is a pattern that is known to be troublesome (hmm, is that what the patterns game is about?).

Thanks for the sample puzzle!

Cheers...

- drac

[gsf]

Not even sure what a low ER variant puzzle is. I've seen some of the submissions but haven't read the rules. Must admit I am less interested in entering a puzzle than I am in finding out techniques (short of recursion, which trivially cracks the puzzle, or educated guessing which I used when I was stuck) that might crack it and if this was/is a pattern that is known to be troublesome (hmm, is that what the patterns game is about?).

the patterns game uses sudoku explainer (SE) to rate puzzles
the rating is a triple ER/EP/ED

ER (explainer rating) the higest SE rating of the methods leading to the puzzle solution
EP (pearl rating) the higest SE rating of the methods leading to the first cell placement
ED (diamond rating) the higest SE rating of the methods leading to the first candidate elimination

in general a lower ER/EP/ED means easier

the game has many goals depending on the participants
but the results of each game are useful for the comminuty as a whole:
a range of ER/EP/ED for a given pattern

for each solver program a certain ER range will roughly represent the knee where the
solver methods (not including guessing/backtracking) fail to solve puzzles
for my solver its 7.3 - 9.3

you don't have to play the game to make a submission
if you submit the low ER variant I posted, at the end of that game you will have access
to a set of puzzles with a range of difficulties
somewhere in that range your solver will start having trouble
and you can look at the puzzles around that point to see how to advance your solver

[champagne]

[quote="gsf]
somewhere in that range your solver will start having trouble
and you can look at the puzzles around that point to see how to advance your solver[/quote]

I got to read your post.

1) I have seen the game and I find it interesting although dedicated (to my view) to puzzles providers, what I am not.

2) What if your solver does not start having trouble??

[Glyn]

Gerard

2) What if your solver does not start having trouble??

I think Glenn is referring to the threshold that exists beyond Grouped AIC. Here SE switches to net like methods. Your program is already capable of progress on this level, but many programs either don't handle them at all or can't describe them in a readily understandable fashion. The subject of this thread has an ER of 9.1.

[gsf]

2) What if your solver does not start having trouble??

trouble is subjective
start a thread on guessing to see how quickly and how much trouble you can get into
(that's a joke, don't really do it, but if you still can't resist, do it on the eureka forum)

I draw the line at methods that make moves (placements/eliminations) without reasons to back up those moves
backtracking fits that bill, as does the P method in my solver, which makes guesses based on candidate degree
(where bivalue/bilocation is degree 2), to see what happens across multiple guesses
basically, if doing the method manually with pencil and paper would require either an eraser or multiple copies of the puzzle, then its guessing
these two methods are in contrast to say, singles, where moves can be made without the fear of them being wrong

[champagne]

2) What if your solver does not start having trouble??

trouble is subjective
start a thread on guessing to see how quickly and how much trouble you can get into
(that's a joke, don't really do it, but if you still can't resist, do it on the eureka forum)

I draw the line at methods that make moves (placements/eliminations) without reasons to back up those moves
backtracking fits that bill, as does the P method in my solver, which makes guesses based on candidate degree
(where bivalue/bilocation is degree 2), to see what happens across multiple guesses
basically, if doing the method manually with pencil and paper would require either an eraser or multiple copies of the puzzle, then its guessing
these two methods are in contrast to say, singles, where moves can be made without the fear of them being wrong

I have to correct my point 2. Should have been at least

2) What if your solver does not start having trouble??

Regarding the rating, I stayed out of discussions mainly because I am fully in line with glyn view focusing more on threshold (wide classes) than on fine ranking. Same for the game you mention; the rule must be clear (and stable), that's it.

I am more in trouble with rating of "hardest puzzles".

I am close to finalize a "relatively simple" solution for metcalf rated 11.4. (somethng a player could have found, although I am deriving the solution from the proposal of my solver).


But I would agree that "hardest puzzles" is a tiny and very specific corner in the SUDOKU field.

[Draco]

the patterns game uses sudoku explainer (SE) to rate puzzles
the rating is a triple ER/EP/ED

ER (explainer rating) the higest SE rating of the methods leading to the puzzle solution
EP (pearl rating) the higest SE rating of the methods leading to the first cell placement
ED (diamond rating) the higest SE rating of the methods leading to the first candidate elimination

in general a lower ER/EP/ED means easier

the game has many goals depending on the participants
but the results of each game are useful for the comminuty as a whole:
a range of ER/EP/ED for a given pattern

for each solver program a certain ER range will roughly represent the knee where the
solver methods (not including guessing/backtracking) fail to solve puzzles
for my solver its 7.3 - 9.3

you don't have to play the game to make a submission
if you submit the low ER variant I posted, at the end of that game you will have access
to a set of puzzles with a range of difficulties
somewhere in that range your solver will start having trouble
and you can look at the puzzles around that point to see how to advance your solver

Now I get it... thanks Glenn! Not sure if I am ready for that much data but I will keep the option in mind.

Cheers...

- drac

[gsf]

Draco, here are the patterns game results for your puzzle
patterns game 0035
I went ahead and entered it, attributed to you
feel free to submit more entries yourself -- you don't have to play to submit

[Draco]

Cool Glenn, thanks!

- drac