The hardest sudokus (new thread)

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

Re: The hardest sudokus (new thread)

Postby Obi-Wahn » Sat Apr 20, 2013 9:44 am

champagne wrote:Congratulations, as a first post, you entered what was considered one year ago as an extremely difficult field.


To be honest, I'm not that new to this forum. I posted as Obi-Wahn back in 2008 but since the forum was moved I can't seem to log into my old account. I even posted three of my own puzzles OW081006, OW081007 and OW081008 with SE 11.4 - 11.7 in the old "The hardest sudokus" thread, which unfortunately never made it to your database and are now attributed to someone else.

Regarding the multifish pattern, it also appears in the other three SE 11.8 puzzles next to the one I discussed above:

Code: Select all
68.7.....3..89......1..4...7...3.9....6...7....2....51.6..8.3.......5.4.........2;11.80;1.50;1.50;225578;col;2012_11_10;0
68.7.....3..89......5..4...7...3.9....6...7....2....51.6..8.3.......1.4.........2;11.80;1.50;1.50;225579;col;2012_11_10;0
........1.....2.3....14.5....3.....6.5.7..4..8.......2..6..8..5.4.51.....9..7.1..;11.80;11.80;2.60;225581;dob;2012_11_10;0

For the last one the digit set would be 2,3,6,8.

BTW, the SK loop can also be described as a rank 0 logic with 16 cell contstraints as base and 4 boxes, 2 rows and 2 columns for 2 digits each as cover. I don't know if it was ever described like that.
In David's example that would be:
base: 2n2346, 7n2346, 1389n1, 1389n5 = 16 truths
cover: 15b1, 37b2, 37b7, 15b8, 48r2, 68r7, 48c1, 26c5 = 16 links
User avatar
Obi-Wahn
 
Posts: 63
Joined: 05 January 2007
Location: Darmstadt, Germany

Re: The hardest sudokus (new thread)

Postby champagne » Sat Apr 20, 2013 3:10 pm

swu wrote:I even posted three of my own puzzles OW081006, OW081007 and OW081008 with SE 11.4 - 11.7 in the old "The hardest sudokus" thread, which unfortunately never made it to your database and are now attributed to someone else.

the general idea is to keep in the data base the oldest reference. I'll update the data base ASAP. (The ownership in not a key point).
swu wrote:Regarding the multifish pattern, it also appears in the other three SE 11.8 puzzles next to the one I discussed above:

I'll work to have an update of exotic properties immediately after the next update of the data base;
swu wrote:BTW, the SK loop can also be described as a rank 0 logic with 16 cell constraints as base and 4 boxes, 2 rows and 2 columns for 2 digits each as cover. I don't know if it was ever described like that.

It is described in the thread "exotic patterns a resume". BTW, it is my preferred rank 0 logic just because it follows exactly the logic I apply to find the SK loop.
champagne
2017 Supporter
 
Posts: 7332
Joined: 02 August 2007
Location: France Brittany

Re: The hardest sudokus (new thread)

Postby champagne » Sat Apr 20, 2013 5:02 pm

Hi Obi_Wahn,

I checked the status of your 3 puzzles.

They came again last year in my own search, so, it will be very easy to change the ownership.
This will be done in the next update
champagne
2017 Supporter
 
Posts: 7332
Joined: 02 August 2007
Location: France Brittany

Re: The hardest sudokus (new thread)

Postby camb » Wed Jun 19, 2013 5:21 pm

Hi Tarek,

I was curious about whether some of the most difficult puzzles were machine generated. For example, I'm using a constraint-based search and find 00041 (ranked in q1's top five) to be exceptionally difficult:

"Label: HardestSudokusThread-00041;tarx0001;Fata_Morgana"

Do you know whether this puzzle was handcrafted or machine generated?

Thanks,
Cameron
camb
 
Posts: 1
Joined: 19 June 2013

Re: The hardest sudokus (new thread)

Postby champagne » Thu Jun 20, 2013 6:41 am

camb wrote:Hi Tarek,

I was curious about whether some of the most difficult puzzles were machine generated. For example, I'm using a constraint-based search and find 00041 (ranked in q1's top five) to be exceptionally difficult:

"Label: HardestSudokusThread-00041;tarx0001;Fata_Morgana"

Do you know whether this puzzle was handcrafted or machine generated?

Thanks,
Cameron


Hi Cameron,

I let tarek answer for "fata morgana" (may be a pm would be better to be sure to touch him but you have to post more than 3 messages to be authorised to use a pm).

Generally speaking, the best people have done by hand IMO is to select a pattern. All the rest is computer work with several tasks

- generate a valid puzzle
- check it is a minimal one
- rate him

The tools vary, but the most commonly used has been the vicinity search to create new puzzles.
Several ratings have been used, none has been really convincing as giving the right ranking. The data base of potential hardest is based on Sudoku Explainer ratings and "fata morgana" has not the necessary rating to enter the data base (and we know how to solve it "easily").


You seem to have a puzzle generator using internal filters. One interesting point, if you are looking for hard puzzles, is the throughput you can achieve with your process (which is not easy to define) compared to others.
champagne
2017 Supporter
 
Posts: 7332
Joined: 02 August 2007
Location: France Brittany

Re: The hardest sudokus (new thread)

Postby tarek » Thu Jun 20, 2013 9:39 am

Hi Cameron & thanks Champagne for your addition.

The puzzle in question is machine generated but not completely. In that period of Hardest puzzles search; the strategy involved choosing a good starting puzzle & do a vicinity search, this search can be stopped at some stage if the results are fruitless with a different start(seed) puzzle. Finding these puzzle at random (ecven with a machine) is almost impossible. Handcrafting these puzzles from scratch is futile.

Tarek
User avatar
tarek
 
Posts: 3762
Joined: 05 January 2006

Re: The hardest sudokus (new thread)

Postby champagne » Thu Jul 18, 2013 8:17 am

I tried to make an update of the potential hardest data base in the skfr project as usual.

Google has discarded that process, that should work till Jan 2014 where it existed.

After several failure, I moved to the new process and loaded the file in my "google drive space"


the link to that storage place is here

The file name is pot_hardest_13_07.zip

This is just the update of the puzzles. I do my best to update in the near future all files of puzzles with exotic properties
champagne
2017 Supporter
 
Posts: 7332
Joined: 02 August 2007
Location: France Brittany

Re: The hardest sudokus (new thread)

Postby champagne » Sat Jul 20, 2013 6:51 am

I started the update of the exotic properties for the potential hardest file.
all new files are in my "google drive space"

the link to that storage place is here

The file "file names phxx.txt" contains the names of the files already loaded or to come.

I am running a new complete check of the r0 logic using the last update of my code.
I am also running a cells-base analysis , but I am not expecting results before about one month. The process is too low.

Meantime, I'll launch a revision of the multi fish potential which is one way to detect puzzles of interest and has a runtime not too high
champagne
2017 Supporter
 
Posts: 7332
Joined: 02 August 2007
Location: France Brittany

Re: The hardest sudokus (new thread)

Postby champagne » Wed Oct 02, 2013 8:52 am

Many things happened during these months and meantime, the generation produced new potential hardest.

I just made an update of the data base and uploaded the new status of the base in the file ph_13_10. Updates of files containing exotic properties will come slowly without notice. Corresponding ph_13_07 files will be erased.


all new files are in my "google drive space"

the link to that storage place is here


here an abstract of the base-status.txt file contained in the ph_13_10.zip file

Code: Select all
Changes in that update are mainly the new puzzles I found mostly in the 25 clues area, but with an entry in the 26 clues area and some puzzles with less clues.

The file contains currently 1096206 puzzles with about 100 000 new entries,
out of which 267821 are rated 11 and more.

The major fact here is the start of a systematic search in the 26 clues area.

An update of the exotic properties will come soon.

the count per number of clues is the following

clues   nb      previous
20   262      260
21   10525      9639
22   74134      68265
23   247220      233499   
24   390692      369237
25   333130      282137
26   40164      663

As one can see, exploration for lower number of clues continues, but the count for 26 clues is now significant, although next months should bring more puzzles.

The 20 clues area seems covered with only 2 new entries. I got more than expected puzzles in tyhe 21_23 area.
The 24 clues area had a poor yield and I have likely to look for fresh seeds in that area. It could be that the area is significantly covered



I prepared a table of high ratings I got so far


   20   21   22   23   24   25   26
11.9      1   4   3         
11.8   1   15   27   7         
11.7   1   30   49   9   10   7   
11.6   5   34   50   41   32   9   
11.5   1   45   98   140   103   57   2
11.4   5   125   299   335   242   114   19
11.3   41   591   1080   1826   1739   894   175
11.2   73   931   2560   4965   5188   2568   456
11.1   61   1213   4717   13999   19383   10104   846
                     
   188   2985   8884   21325   25697   13753   1498

I find that table more and more interesting. Areas 25 clues and 26 clues have still a lot of potential.

For sure, highest ratings seem to be covered, but we have here a kind of numeric bias. The less given you have, the more candidates you should get and a tough clearing move has a good chance to use more candidates, what in serate gives a higher rating.

26 clues's area  seems to be still able to produce many many potential hardest puzzles.
champagne
2017 Supporter
 
Posts: 7332
Joined: 02 August 2007
Location: France Brittany

Re: The hardest sudokus (new thread)

Postby champagne » Fri Dec 06, 2013 4:57 pm

I added a little more than 99 000 puzzles, stored in a separate file of my google drive "php13_11".

As I did not work on the properties of these new puzzles, I only published additions (a kind of safety measure)
I worked again in the 21-22 area, but I'll comment on that in another thread.

Most puzzles are in the 25_26 clues areas
champagne
2017 Supporter
 
Posts: 7332
Joined: 02 August 2007
Location: France Brittany

Re: The hardest sudokus (new thread)

Postby champagne » Tue Dec 31, 2013 9:00 am

champagne wrote:I added a little more than 99 000 puzzles, stored in a separate file of my google drive "php13_11".

As I did not work on the properties of these new puzzles, I only published additions (a kind of safety measure)
I worked again in the 21-22 area, but I'll comment on that in another thread.

Most puzzles are in the 25_26 clues areas


New update in the same conditions. the last file php_14_01 has 135222 puzzles. The file php13_11 has been cancelled
A significant part of that lot is in the field 22 clues and is derived form the "exocet pattern" generation.
Some puzzles have been added by mladen dobrichev.

Next update of the data base will not come before end of march.


Happy new year to all
champagne
2017 Supporter
 
Posts: 7332
Joined: 02 August 2007
Location: France Brittany

Re: The hardest sudokus (new thread)

Postby dobrichev » Tue Dec 31, 2013 2:01 pm

The latest collection is the union of ph_13_10 and php_14_01 files, right?

If so, there are 8876 duplicated puzzles.
The number of unique puzzles is 1231428.

Below is the distribution by solution grids.
Hidden Text: Show
Code: Select all
 #grids #puzzles
      1 651
      1 640
      1 553
      1 488
      1 479
      1 458
      1 450
      1 411
      1 410
      1 408
      1 392
      1 376
      1 373
      1 360
      1 318
      1 316
      1 306
      1 224
      1 212
      1 197
      1 181
      1 180
      1 171
      1 165
      2 154
      1 153
      1 137
      1 134
      1 132
      1 131
      1 129
      1 128
      1 116
      1 111
      3 107
      1 106
      2 100
      3 98
      2 97
      1 96
      2 94
      1 93
      2 91
      1 89
      2 88
      1 86
      3 82
      6 81
      3 80
      4 79
      2 78
      5 77
      5 76
      2 75
      5 74
      3 73
      4 71
      4 70
      5 69
      2 68
      3 67
      2 66
      5 65
      4 64
      3 63
      7 62
      7 61
      1 60
      1 59
      9 58
      4 57
      6 56
      7 55
      4 54
      6 53
      8 52
     13 51
      8 50
     10 49
     13 48
      6 47
      7 46
     17 45
     11 44
     17 43
     13 42
     13 41
     15 40
     17 39
     26 38
     24 37
     24 36
     38 35
     37 34
     35 33
     36 32
     28 31
     36 30
     44 29
     60 28
     69 27
     65 26
     63 25
     82 24
     94 23
    100 22
    113 21
    139 20
    128 19
    202 18
    208 17
    269 16
    256 15
    392 14
    468 13
    609 12
    728 11
   1087 10
   1289 9
   2098 8
   2614 7
   4720 6
   7237 5
  15594 4
  32347 3
 107724 2
 624020 1

Every second puzzle has brothers born in the same solution grid family.

Pattern distribution looks in similar way
Hidden Text: Show
Code: Select all
#patterns #puzzles
      1     939
      1     827
      1     809
      1     716
      1     624
      1     559
      1     513
      1     488
      1     486
      1     464
      1     447
      1     415
      1     412
      1     383
      1     350
      1     344
      1     338
      1     330
      1     329
      1     309
      1     307
      1     304
      1     301
      2     300
      1     296
      1     295
      1     293
      1     292
      1     286
      1     279
      1     276
      1     273
      1     272
      1     268
      1     263
      1     255
      1     254
      1     249
      1     248
      1     243
      1     231
      1     224
      1     216
      1     212
      1     211
      1     205
      1     201
      1     198
      1     193
      1     191
      1     187
      1     186
      1     183
      1     180
      3     179
      1     174
      3     169
      2     167
      2     165
      1     164
      1     163
      2     162
      1     161
      1     160
      1     157
      1     153
      1     151
      1     150
      1     148
      1     147
      1     146
      3     144
      2     143
      1     139
      1     138
      2     137
      1     134
      1     131
      1     129
      1     128
      2     127
      1     126
      2     125
      2     124
      1     122
      2     121
      3     120
      3     118
      3     115
      2     114
      1     113
      3     112
      2     111
      2     110
      1     109
      1     108
      1     107
      4     106
      1     105
      1     104
      3     103
      2     102
      1     101
      4     100
      2      99
      1      98
      5      97
      3      96
      3      95
      4      94
      3      93
      2      92
      2      91
      4      90
      3      89
      9      88
      3      87
      5      86
      2      85
      3      84
      3      83
      6      82
      4      81
      3      80
      8      79
      4      78
      3      77
      8      76
      9      75
     10      74
      8      73
      4      72
      8      71
      9      70
      5      69
      6      68
     11      67
      8      66
      8      65
     13      64
     12      63
      7      62
     15      61
     10      60
     12      59
     16      58
     16      57
     13      56
      9      55
     14      54
     15      53
     13      52
     16      51
     15      50
      9      49
     17      48
     21      47
     22      46
     24      45
     22      44
     33      43
     33      42
     35      41
     31      40
     22      39
     38      38
     34      37
     39      36
     41      35
     40      34
     53      33
     70      32
     55      31
     65      30
     84      29
     86      28
    103      27
     84      26
    103      25
    118      24
    120      23
    129      22
    172      21
    200      20
    246      19
    257      18
    304      17
    349      16
    433      15
    548      14
    651      13
    763      12
    979      11
   1281      10
   1732       9
   2550       8
   3462       7
   5462       6
   8328       5
  17000       4
  31970       3
 110029       2
 518959       1
dobrichev
2016 Supporter
 
Posts: 1843
Joined: 24 May 2010

Re: The hardest sudokus (new thread)

Postby champagne » Tue Dec 31, 2013 3:41 pm

dobrichev wrote:The latest collection is the union of ph_13_10 and php_14_01 files, right?

If so, there are 8876 duplicated puzzles.
The number of unique puzzles is 1231428.



The full collection should be the union and the data base contains also 1231428 puzzles.

I surely made a mistake when I entered the filter for the complementary file.

I'll check that in March but at that time, I'll try to post a new full analysis
champagne
2017 Supporter
 
Posts: 7332
Joined: 02 August 2007
Location: France Brittany

Re: The hardest sudokus (new thread)

Postby champagne » Sat Apr 05, 2014 9:20 am

I just made an update of the data base as of late march 2014.

as for all new files , that update is in my "google drive space" with the name ph14_04.zip

the link to that storage place is here



The file contains now 1 264 655 puzzles
33 227 puzzles have been added since January, mainly in the 26 clues field.

I did not work since December on the exotic properties. On my side, this will come only in May. I have other constraints in April.

The clue count in the data base was not correct for several puzzles, especially in the 20_22 clues field.
The data base has been adjusted. The statistics are now in line with the adjusted count


the count per clue in the data base is the following

Code: Select all
20   77
21   10791
22   86274
23   254119
24   413391
25   355401
26   144419
more   183


Interesting is the distribution of highest ratings per clue.
The 26 clues field is far from being investigated (same somehow at a lowest degree for 24 and 25 clues),
but we can already have some feelings

Code: Select all
          20     21    22      23      24      25      26
11.9              1     4       3         
11.8       1     15    26       8         
11.7             31    50       8      13       8      
11.6       1     38    50      40      34      10       1   
11.5             46    99     138     102      75      12   
11.4            130    30     332     251     156      49   
11.3       1    627  1102    1763    1818    1055     539   
11.2       2    989  2670    4900    5376    2839    1176   
11.1      12   1263  4976   13961   18977   11389    3407   
11         3    802  5042   17835   25694   14006    6178                  
                        
          20   3942 14320   38988   52265   29538   11362   150435

Due to the Sudoku Explainer rating rules, we have a direct effect of the number of candidates on the rating, not easy to quantify.

So IMO the results for 26 clues are not so far from the results for lower counts.
champagne
2017 Supporter
 
Posts: 7332
Joined: 02 August 2007
Location: France Brittany

Re: The hardest sudokus (new thread)

Postby petyo staykov » Thu May 15, 2014 11:20 am

here's a new very difficult puzzle. It has ER/EP/ED 11,9,11.9,11.8
..8..31..
.5..7....
2..9.....
9..3.....
.7..2..3.
..3..8..6
...5....4
..4..6.1.
......6.8

Good luck to all members of the forum.
Regards,
Petyo Staykov
petyo staykov
 
Posts: 4
Joined: 11 May 2014

PreviousNext

Return to General