The New Sudoku Players' Forum

[Eioru]

these 4 are another pattern and 22 clue
2..1..7......3......4..5..81..2..6...3.....5...7..4..95..6..8......4......8..9..3
1..4..6......5......5..3..72..6..8...5.....4...3..2..97..1..4......7......9..8..3
4..3..8......4......3..6..78..9..2...7.....1...5..7..39..8..7......2......1..5..6
5..3..7......9......2..8..33..4..5...1.....8...6..7..44..7..9......1......1..2..6

[Eioru]

here are 10 permutations of the 10.0 puzzle
could someone with more patience (or maybe a better jre) report the explainer ratings
thanks

Code: Select all

.....1..2.1..2..3.4..5.......4.....6.7..3..1.8.....9..5....8.......1..7...64..5..
.2....1.......4..53...1..2.1...7..3...9..8....6.4..........58....56....47......1.
...8..5....7.1.....5...4..6..1.3..7..9....8..6.......4..3.2..1.2..1..........54..
4.....5......2...1..13...2..6...54.....7...1.5.......8.4..6......71...3.8....9...
..5.....8.6...54......7..1.7...1..3...8..9....4.6........2....11...3..2...4...5..
.5..8.........17..6..4...5...1..23...4.5.........1...2..7..31...8.....9.4.......6
.7..1.........8..5..54...6..1..3.7....9.....86......4....5....42....1....3..2.1..
4.....6...8......9..73...1..5...8...6...4...5...1...7...12...3..4..5.........12..
5...6..4......58...7......1.3.1....2.....4.5...2...1....6.4.....1.7....39....8...
...5...8..5..6.4....7.....12......1....4..5....3..1..26...4......1..7..3.9.8.....


How difficulty are they (Cause all are 10.0 by SE)
It might be another way to analyese them.

[Eioru]

Find these rated 9.0~9.8 analysed by SE
1..2..3......1......2..4..53..6..7...5.....8...9..5..26..3..5......7......8..9..4
1..2..3......1......4..5..62..7..5...5.....8...7..8..49..3..2......9......6..4..9
1..2..3......1......4..5..63..4..7...8.....2...9..6..57..8..6......3......3..4..9
1..2..3......1......4..5..63..7..8...1.....4...9..8..57..3..6......8......2..4..9
1..2..3......4......2..5..63..7..8...5.....4...7..9..24..8..6......3......9..7..1
1..2..3......4......3..5..62..1..7...8.....9...6..8..44..9..2......8......5..3..7
1..2..3......4......3..5..67..4..6...5.....2...6..8..92..1..8......9......8..6..7
1..2..3......4......4..3..53..6..1...7.....2...8..9..46..7..2......5......9..8..1
1..2..3......4......4..5..26..1..7...8.....4...5..8..97..3..9......1......2..7..5
1..2..3......4......4..5..67..3..8...4.....2...5..7..96..1..2......6......9..8..5
1..2..3......4......4..5..67..3..8...9.....5...6..1..92..9..7......8......5..7..3
1..2..3......4......5..1..62..3..1...7.....8...9..6..24..1..7......9......7..5..9
1..2..3......4......5..1..67..8..4...6.....1...1..9..24..9..2......7......8..5..9
1..2..3......4......5..3..62..1..7...8.....5...4..6..89..7..1......9......3..4..5
1..2..3......4......5..3..62..7..4...8.....1...6..9..59..8..2......7......1..5..4
1..2..3......4......5..3..62..7..4...8.....3...1..9..57..1..8......2......6..5..9
1..2..3......4......5..6..12..7..4...6.....1...4..8..79..3..2......9......7..5..8
1..2..3......4......5..6..22..7..1...8.....6...9..3..77..3..4......8......8..5..9
1..2..3......4......5..6..24..3..7...8.....6...1..5..98..7..2......3......6..9..1
1..2..3......4......5..6..27..8..2...3.....5...4..7..62..1..9......8......9..5..4
1..2..3......4......5..6..43..1..7...2.....8...9..8..56..4..9......7......8..5..7
1..2..3......4......5..6..72..6..8...7.....2...4..8..93..7..6......1......9..5..4
1..2..3......4......5..6..73..6..2...8.....1...7..1..92..8..9......7......6..4..5
1..2..3......4......5..6..73..8..9...8.....7...6..5..12..3..4......6......9..7..8
1..2..3......4......5..6..74..3..8...8.....7...6..9..53..8..9......2......2..5..1
1..2..3......4......5..6..75..1..8...8.....2...6..5..48..9..2......3......7..2..6
1..2..3......4......5..6..75..8..4...9.....1...3..7..64..9..5......8......6..2..9
1..2..3......4......5..6..78..1..2...7.....1...9..8..53..8..5......6......4..5..9
1..2..3......4......5..6..78..7..1...3.....8...6..5..42..6..5......9......4..8..6

[ravel]

Updated the hardest list with puzzles from Eioru's lists:

In the first list (4 puzzles) nr 4(9.8) needed 6 steps, 2 and 3 had 4 steps.

In the second list (29 puzzles) nr 18(9.8) needed 6 steps, nr 12(9.3) 5 steps and numbers 1,3,9,10,18 and 27 had 4 steps.
In this list 3 puzzles are not minimal (lines 4,7 and 21). From the minimal versions (in all three the given in r4c7 can be dropped), the puzzle of line 4 was ultrahard too (5 steps/9.1):

Code: Select all

1..2..3......1......4..5..63..7..8...1.....4...9..8..57..3..6......8......2..4..9
1..2..3......1......4..5..63..7......1.....4...9..8..57..3..6......8......2..4..9

[Eioru]

Updated the hardest list with puzzles from Eioru's lists:

In the first list (4 puzzles) nr 4(9.8) needed 6 steps, 2 and 3 had 4 steps.

In the second list (29 puzzles) nr 18(9.8) needed 6 steps

I normalize all 4 of first list to second list... ( 4 first list can be delete from your hardest list )

and I also find others

1..2..3......1......4..5..67..8..2...4.....9...9..3..13..1..8......7......6..9..5
1..2..3......4......5..1..64..1..2...3.....7...6..8..59..7..4......9......7..6..3
1..2..3......4......5..1..67..8..2...3.....5...4..9..78..6..7......1......9..5..4
1..2..3......4......5..3..64..7..8...3.....5...6..9..77..8..1......1......2..6..9
1..2..3......4......5..3..66..7..2...7.....5...8..9..42..1..9......6......7..4..8
1..2..3......4......5..6..13..7..8...9.....2...4..2..79..1..7......5......6..8..4
1..2..3......4......5..6..45..7..1...4.....6...8..2..37..6..2......9......9..5..8
1..2..3......4......5..6..47..3..1.........6...6..4..83..7..2.............8..9..5
1..2..3......4......5..6..47..3..1...8.........6..4..93..7..2.............9..8..5
1..2..3......4......5..6..73..6..8...4.....5...9..5..68..3..9......7......1..2..4
1..2..3......4......5..6..73..8..1...7.....5...6..7..49..5..8......9......1..4..2
1..2..3......4......5..6..74..6..8...7.....1...9..5..32..8..5......2......6..7..9
1..2..3......4......5..6..77..6..2...8.....9...4..3..59..3..8......5......6..7..1
1..2..3......4......5..6..78..9..2...7.....4...1..8..64..3..8......2......6..5..3

[tarek]

my solver is a bit slow to check larger number of very difficult puzzles, Is there a solver out there that can screen a number of puzzzles qiuckly so that I would then have a smaller number to work on..........

tarek

[ravel]

Don't know, i have the same problem. gsf's solver is very fast for a first rating, but his 8- and 9-rated puzzles still differ very much, when i check the step counts.
Same for Sudoku Explainer - which also is slow and i did not find a way so far, how i could automize to let it rate a list over night.
Ocean's last list did not contain a single puzzle, which did not qualify for the list, but i think he also had to put a lot of work and cpu time in putting it together.

[gsf]

Don't know, i have the same problem. gsf's solver is very fast for a first rating, but his 8- and 9-rated puzzles still differ very much, when i check the step counts.

in general 8 means at least one backdoor single, 9 means no backdoor single and at least one backdoor pair

rating variations for harder puzzles is maddening
in my case its due to the luck of the backtrack guesses when the non-backtracking constraints fail to progress the solution
the guessing is normalized a bit by checking cells with less candidates first

I'm looking at using error net stats to classify unconstrained ("harder") sudoku
it looks at the total number of error net propositions vs the number of contradictions and solutions uncovered

I'm wrestling with the formalization now -- it involves:
the solution constraint set (singles, pairs, x-wing etc.)
the constraints propagated in the error net (usually just singles)
the girth (or depth) of the error net propagations (i.e., the proposition ripple effect)
batching the error net moves at one position until all are identified

[tarek]

I don't mind the explainer, I just get very annoyed when I need to manually check each puzzle, I could fairly quickly check for (7.5+ explainer rating) using my solver , I'm happy to go through a number <50 but with larger numbers, if there is a way to shortlist 50 out of 1000 puzzles scoring (7.5+ explainer rating), then I would be happy, I found several between 9.0-9.3 I will post anything that scores 9.8 or greater

tarek

[ravel]

I'm looking at using error net stats to classify unconstrained ("harder") sudoku

Glad to hear that, i am a fan of your software

[daj95376]

I suggest that you concentrate on puzzles #1 and #9. After that, they're about the same in difficulty for my solver.
(Note: none need complex elimination chains.)

[ravel]

In Eioru's last list my program rated numbers 1,2,6,10,13 and 14 with 4 steps.
Number 9 was resolved with 3 steps: r2c6<>1, r2c3<>3, r1c6<>7 (it stops after finding the first solution with less than 4 steps).

[tarek]

This is the first time that I've tried to see different solvers from the internet.......

For the purpose of this thread: Not that many can handle multiple puzzle at once & give a rating..........

gsf's sudoku looks promising......it is fast, however as you said ravel it doesn't give a real indication for the extreme hard ones...... I used the following parameters -f%Xr Input>Output, I might be using the wrong parameters.

I get 9/21/00/0349/2 for one of the puzzles that corresponds to 9.2 explainer..........

tarek

[Ocean]

Ocean's last list did not contain a single puzzle, which did not qualify for the list

Seems that puzzles in those patterns with ER 9.2 or above have high possibility of making it to the list. In the fully symmetric 20 pattern some puzzles with 9.2 did not qualify. Puzzles with ER 9.1, 9.0 or lower may qualify, but several do not.

Is there a solver out there that can screen a number of puzzles quickly so that I would then have a smaller number to work on

As a first screening tool I have used a method similar to dukuso's "suexrat". Ordering the puzzles based on their "suexrating", starting with the highest and testing them in Explainer one by one. No perfect method, but it did end up with some high-rated puzzles.


I have also used gsf's program (sudoku) as a second screening tool from time to time. It will be good for initial screening also. But my puzzles are generated and live on a different machine, and I don't know how to transfer them effectively (a diskette only takes about 15000 puzzles, I don't like to burn new CDs all the time, and mail is no good option either. Maybe I could buy one of those memory pens. But I prefer screening close to the source, not having to transfer large bulks of data.)

[gsf]

This is the first time that I've tried to see different solvers from the internet.......

For the purpose of this thread: Not that many can handle multiple puzzle at once & give a rating..........

gsf's sudoku looks promising......it is fast, however as you said ravel it doesn't give a real indication for the extreme hard ones...... I used the following parameters -f%Xr Input>Output, I might be using the wrong parameters.

I get 9/21/00/0349/2 for one of the puzzles that corresponds to 9.2 explainer..........

try -f%r for ratings in the range 0..99999, the 90000..99999 range is exponential
see --man for a description of -R that lets you change the rating expression
the expression is based on <char><digit> identifiers that access the stats
collected for the puzzle solution

the stats are biased by the luck of the backtrack search for the hard puzzles
I'm close to removing that bias