The New Sudoku Players' Forum

[dobrichev]

Thank you.

To get a more accurate figure, one should apply an enumeration process.

This is what I am playing with.
I ran a new (and unstable) enumerator for ~20 hours and it generated ~8 millions puzzles not taking into account the symmetry, i.e about 4 millions ED puzzles. That is 40 days for 200 millions puzzles, reducible by symmetry and parallelisation.
There are bugs which, when fixed, can change the performance in orders of magnitude.
The same code enumerates a 17-given pattern for 2 to 3 hours.

[champagne]

Thank you.

To get a more accurate figure, one should apply an enumeration process.

This is what I am playing with.
I ran a new (and unstable) enumerator for ~20 hours and it generated ~8 millions puzzles not taking into account the symmetry, i.e about 4 millions ED puzzles. That is 40 days for 200 millions puzzles, reducible by symmetry and parallelisation.
There are bugs which, when fixed, can change the performance in orders of magnitude.
The same code enumerates a 17-given pattern for 2 to 3 hours.

Hi Mladen,

I was thinking of a statistical approach on a small sample.

If I got it, you are doing a full scan. This has always been on my side a very slow process compared to the vicinity search.

I generated the 100 millions puzzles in about 3 days using an i7.

[dobrichev]

My focus is on exhaustive enumerator and obviously it wouldn't work for patterns game today. I want to estimate the time needed for enumeration of patterns of reasonable size.

[coloin]

Any idea about the total number of ED puzzles in this pattern?

always difficult to answer to such a question out of realized figures, but I have 100 millions ED puzzles in my data base, so the minimum should be 200 millions.
I likely covered more than 80% of the high ratings, but the yield of my search is much lower in the low ratings.

To get a more accurate figure, one should apply an enumeration process.

Well here I estimated that there would be 18 billion ED puzzles per 25 clue pattern.
A symmetrical pattern could be reduced by a factor.
But as ever the permutations are large - and possibly an underestimate
C

[champagne]

Hi Coloin,

I have nothing to object to your 18 billions compared to my poor estimate of 200 millions. I believe that you have the best estimate. I would be interested in the estimate of the high ratings, but this seems very difficult to do, especially considering the huge differences from one pattern to another one.

In the pattern game, the bottleneck is definitely the rating; I can surely improve skfr using the last findings in the brute force, but not that much.

Anyway, over 200 millions puzzles, the chances to find a new submission are close to null, and I see no other incentive to explore deeply a specific pattern.

The search of potential hardest remains the main reason on my side to speed up the rating program.

[g.r.emlin]

::: game 0260 start LIGHTNING at 2015-10-12+23:00:00-0000 :::

Code: Select all

dealer status

candidate patterns (via private message to g.r.emlin)

   4 m_b_metcalf      17 champagne        87 g.r.emlin

past patterns

   1 claudiarabia      1 ravel             9 gsf              30 Mauricio         58 tarek
   1 draco             1 wintder          19 TTHsieh          32 m_b_metcalf
   1 Eioru             3 g.r.emlin        21 coloin           33 JPF
   1 pseudocoup        6 ronk             26 champagne        47 Pat


[g.r.emlin]

::: game 0260 open LIGHTNING 2015-10-12+23:00:08-0000 dealer champagne 23 clues pi rotational (180 degrees) symmetry :::

Code: Select all

3 . . 2 . . 9 . .
. 8 . . 1 . . 4 .
. . 1 . . 6 . . .
4 . 5 . . . . 9 .
. . . . 7 . . . .
. 9 . . . . 3 . 5
. . . 8 . . 5 . .
. 2 . . 9 . . 6 .
. . 4 . . 3 . . 1  ED=1.7/1.2/1.2

[CSC]

Code: Select all

1..2..3..
.2..4..5.
..6..7...
7.4....8.
....8....
.8....9.5
...4..1..
.7..9..2.
..3..2..6  ED=9.2/9.2/9.2

[m_b_metcalf]

Code: Select all

 1 . . 2 . . 3 . .
 . 4 . . 5 . . 6 .
 . . 7 . . 8 . . .
 5 . 8 . . . . 7 .
 . . . . 4 . . . .
 . 3 . . . . 9 . 5
 . . . 3 . . 2 . .
 . 6 . . 1 . . 4 .
 . . 9 . . 7 . . 8   ED=9.1/9.1/9.1

[CSC]

Code: Select all

1..2..3..
.4..5..2.
..6..7...
8.4....9.
....3....
.9....5.1
...1..2..
.8..4..5.
..3..6..7  ED=9.0/9.0/9.0

[m_b_metcalf]

Code: Select all

 1 . . 2 . . 3 . .
 . 3 . . 1 . . 4 .
 . . 5 . . 6 . . .
 7 . 6 . . . . 8 .
 . . . . 2 . . . .
 . 9 . . . . 1 . 7
 . . . 1 . . 8 . .
 . 4 . . 9 . . 3 .
 . . 8 . . 5 . . 6   ED=9.3/9.3/9.3

[champagne]

Code: Select all

1..2..3..
.4..5..6.
..2..7...
2.4....8.
....8....
.8....5.6
...9..1..
.2..4..5.
..7..3..9 ED=10.4/10.4/10.4

[m_b_metcalf]

Code: Select all

 1 . . 2 . . 3 . .
 . 4 . . 5 . . 6 .
 . . 7 . . 8 . . .
 2 . 9 . . . . 5 .
 . . . . 3 . . . .
 . 8 . . . . 1 . 9
 . . . 5 . . 2 . .
 . 6 . . 8 . . 4 .
 . . 5 . . 7 . . 8   ED=9.4/9.4/9.4

[CSC]

Code: Select all

1..2..3..
.4..5..6.
..7..8...
2.4....1.
....9....
.9....4.7
...3..2..
.5..6..4.
..8..7..1  ED=8.5/8.5/8.5

[m_b_metcalf]

Code: Select all

 1 . . 2 . . 3 . .
 . 4 . . 5 . . 6 .
 . . 7 . . 8 . . .
 7 . 4 . . . . 9 .
 . . . . 9 . . . .
 . 9 . . . . 5 . 6
 . . . 3 . . 1 . .
 . 7 . . 4 . . 5 .
 . . 8 . . 5 . . 2   ED=9.8/9.8/9.8