The New Sudoku Players' Forum

[champagne]

Code: Select all

.9......8
7....6...
....5.4..
.....2.1.
..8.4.7..
.1.9.....
..4.6....
...8....1
5......9.  ED=6.6/6.6/3.4

[ano1]

::: comment :::

It looks like the game could end in about 11 hours ...

You guys have done well

I finished my puzzle search some time ago. The preliminary ratings pass on the (symmetric-)minlex puzzles finished about an hour after champagne's last entry.

I am rating the other 3 "same shape"/"no-digit-remap" isomorphs of the puzzles. It will take another 23-24 hours to complete. It looks like there will be 150-200 puzzles that are what I'm calling "SE anomalies" ... where the ratings aren't the same for all isomorphs.

I should mention that I really don't know how well SE handles puzzles that differ only by a digit remapping. I tried a few of the anomolous cases I had seen, and it didn't seem to matter. I didn't try very hard, though -- two cases for each isomorph, for a handful of cases.

The puzzle search produced 21089 puzzles, 367 were non-minimal, leaving 20722 minimal puzzles.
I don't want to say more than that, until the game closes.

I had in mind to post a list of ratings, with puzzle counts for each, if it would be interesting -- something similar to the "game status" posts.
My guess is that you guys covered most if not all of them -- ratings and puzzles !

Here's something interesting: (only) two of the puzzles are automorphic.

Code: Select all

. 1 . . . . . . 2
3 . . . . 4 . . .
. . . . 3 . 5 . .
. . . . . 6 . 3 .
. . 1 . 5 . 7 . .
. 8 . 2 . . . . .
. . 5 . 7 . . . .
. . . 1 . . . . 8
6 . . . . . . 4 .

     - automorphism: diagonal reflection + digit remapping
     - it's minmal
     - it's an 8-digit puzzle


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

     - automorphism: 180 degree rotation + digit remapping
     - it's non-minimal
     - (only) the 7 in r5c5 is redundant
     - removing the 7, leaves another 8-digit puzzle


.1......23....4.......3.5.......6.3...1.5.7...8.2.......5.7.......1....86......4.
.1......23....4.......3.5.......6.3...2.7.6...8.2.......9.8.......1....86......4.


BTW, there are only 27 puzzles with the "18" shape for this game. There are 221 for the 18 related to the game 32 shape. The 221 number is probably the highest count for DA-18's. (I had 199 of them, before the count ... which was why I was interested in it). The "runner up" had 179 puzzles.

Mike, I liked your X's ... here and on the 18's thread. I had never tried solving a 9x9 X before. I got the easy ones, but got stumped early on the hard ones. I'm not much of puzzle solver. Please tell me they needed something beyond locked sets -- finless fish at least

Cheers,
Ano1.

[JPF]

::: comment :::

Thanks ano1.

I turned my computer off when champagne pinched my last potential submission ( i.e. 6.6/6.6/3.4).
"almost randomly", I got 17784 minimal puzzles, including the automorphic one

JPF

[g.r.emlin]

::: game 0103 idle 2010-03-06+13:42:10-0000 -- with no further entries the game will close at 2010-03-06+16:00:00-0000 :::

Code: Select all

game 0103 2010-03-02+12:00:00-0000 duration 2d06h dealer m_b_metcalf pages 468-477 (open) 6 players
________________________________________________________________________________________________________________

    1.2/1.2/1.2 m_b_metcalf    4.0/1.2/1.2 gsf            6.7/3.4/3.4 champagne      8.0/1.2/1.2 JPF
    1.5/1.2/1.2 m_b_metcalf    4.2/3.4/3.4 champagne      6.8/3.4/3.4 gsf            8.2/8.2/3.8 champagne
    1.7/1.2/1.2 m_b_metcalf    4.4/1.2/1.2 m_b_metcalf    6.9/1.2/1.2 JPF            8.3/8.3/3.8 m_b_metcalf
    2.0/1.2/1.2 champagne      4.5/3.8/3.4 JPF            7.0/1.2/1.2 JPF            8.4/8.4/3.8 m_b_metcalf
    2.3/1.2/1.2 m_b_metcalf    4.6/3.4/3.4 m_b_metcalf    7.1/7.1/3.8 champagne      8.5/8.5/3.8 champagne
    2.5/1.2/1.2 gsf            4.7/1.2/1.2 gsf            7.2/7.2/3.8 joel64         8.6/8.6/3.8 gsf
    2.6/1.2/1.2 m_b_metcalf    5.0/1.2/1.2 m_b_metcalf    7.3/7.3/3.8 m_b_metcalf    8.7/8.7/3.4 JPF
    2.8/1.2/1.2 m_b_metcalf    5.2/3.4/3.4 champagne      7.4/7.4/3.4 JPF            8.8/8.8/3.8 joel64
    3.0/1.2/1.2 joel64         5.6/3.4/3.4 champagne      7.5/7.5/3.4 m_b_metcalf    8.9/8.9/3.8 champagne
    3.2/1.2/1.2 m_b_metcalf    5.7/3.4/3.4 JPF            7.6/7.3/3.4 gsf            9.0/9.0/3.8 m_b_metcalf
    3.4/3.4/3.4 gsf            6.2/1.2/1.2 gsf            7.7/7.3/3.4 JPF            9.1/9.1/3.4 JPF
    3.6/1.2/1.2 m_b_metcalf    6.5/1.2/1.2 JPF            7.8/7.8/3.8 champagne      9.2/9.2/3.4 JPF
    3.8/3.8/3.8 gsf            6.6/6.6/3.4 champagne      7.9/1.2/1.2 JPF            9.3/9.3/3.4 m_b_metcalf
________________________________________________________________________________________________________________

    open ratings : 4.8 4.9 5.1 5.4 5.8-6.1 8.1 9.4-11.4

       submitter  entries  score  response  notes
     -----------  -------  -----  --------  -----
             Pat        0      0    22m59s
          joel64        3    203    20m35s  rare-0,pearl*2,premature-entry
             gsf        9    902     1h17m  rare-0*2,rare-1*2,rare-2*2,diamond*2,pearl
             JPF       12   1160     1h33m  rare-0,rare-1,pearl*4
       champagne       11   1273    27m02s  rare-0*4,rare-1*2,rare-2*2,pearl*6,last-entry=226,dup-puzzle,dup-rating
     m_b_metcalf       17   1374    14m47s  rare-0*6,rare-1*2,pearl*6,highest-rating=152,dealer
     -----------  -------  -----            -----
               6       52   4912            rare-0*14,rare-1*7,rare-2*4,diamond*2,pearl*19


[g.r.emlin]

::: game 0103 closed (idle for 1d21h) -- preliminary results :::

Code: Select all

game 0103 2010-03-02+12:00:00-0000 duration 2d06h dealer m_b_metcalf pages 468-477 (final) 6 players
________________________________________________________________________________________________________________

    1.2/1.2/1.2 m_b_metcalf    4.0/1.2/1.2 gsf            6.7/3.4/3.4 champagne      8.0/1.2/1.2 JPF
    1.5/1.2/1.2 m_b_metcalf    4.2/3.4/3.4 champagne      6.8/3.4/3.4 gsf            8.2/8.2/3.8 champagne
    1.7/1.2/1.2 m_b_metcalf    4.4/1.2/1.2 m_b_metcalf    6.9/1.2/1.2 JPF            8.3/8.3/3.8 m_b_metcalf
    2.0/1.2/1.2 champagne      4.5/3.8/3.4 JPF            7.0/1.2/1.2 JPF            8.4/8.4/3.8 m_b_metcalf
    2.3/1.2/1.2 m_b_metcalf    4.6/3.4/3.4 m_b_metcalf    7.1/7.1/3.8 champagne      8.5/8.5/3.8 champagne
    2.5/1.2/1.2 gsf            4.7/1.2/1.2 gsf            7.2/7.2/3.8 joel64         8.6/8.6/3.8 gsf
    2.6/1.2/1.2 m_b_metcalf    5.0/1.2/1.2 m_b_metcalf    7.3/7.3/3.8 m_b_metcalf    8.7/8.7/3.4 JPF
    2.8/1.2/1.2 m_b_metcalf    5.2/3.4/3.4 champagne      7.4/7.4/3.4 JPF            8.8/8.8/3.8 joel64
    3.0/1.2/1.2 joel64         5.6/3.4/3.4 champagne      7.5/7.5/3.4 m_b_metcalf    8.9/8.9/3.8 champagne
    3.2/1.2/1.2 m_b_metcalf    5.7/3.4/3.4 JPF            7.6/7.3/3.4 gsf            9.0/9.0/3.8 m_b_metcalf
    3.4/3.4/3.4 gsf            6.2/1.2/1.2 gsf            7.7/7.3/3.4 JPF            9.1/9.1/3.4 JPF
    3.6/1.2/1.2 m_b_metcalf    6.5/1.2/1.2 JPF            7.8/7.8/3.8 champagne      9.2/9.2/3.4 JPF
    3.8/3.8/3.8 gsf            6.6/6.6/3.4 champagne      7.9/1.2/1.2 JPF            9.3/9.3/3.4 m_b_metcalf
________________________________________________________________________________________________________________

    open ratings : 4.8 4.9 5.1 5.4 5.8-6.1 8.1 9.4-11.4

       submitter  entries  score  response  notes
     -----------  -------  -----  --------  -----
             Pat        0      0    22m59s
          joel64        3    203    20m35s  rare-0,pearl*2,premature-entry
             gsf        9    902     1h17m  rare-0*2,rare-1*2,rare-2*2,diamond*2,pearl
             JPF       12   1160     1h33m  rare-0,rare-1,pearl*4
       champagne       11   1273    27m02s  rare-0*4,rare-1*2,rare-2*2,pearl*6,last-entry=226,dup-puzzle,dup-rating
     m_b_metcalf       17   1374    14m47s  rare-0*6,rare-1*2,pearl*6,highest-rating=152,dealer
     -----------  -------  -----            -----
               6       52   4912            rare-0*14,rare-1*7,rare-2*4,diamond*2,pearl*19


[ano1]

::: comment :::

One "::: more :::" entry from champagne or JPF, and I'll holler "bingo".

[champagne]

:::comment:::

::: comment :::

One "::: more :::" entry from champagne or JPF, and I'll holler "bingo".

Sorry, but as JPF, I turned the computer off one day ago.

On top of it, contrary to JPF, I run the game using filters, so I did not generate all the puzzles in the field I covered.

I rated less than 5000 puzzles in that game. Nevertheless, I have nearly all the final entries in my cache.

champagne

[ano1]

::: comment :::

I turned my computer off when champagne pinched my last potential submission ( i.e. 6.6/6.6/3.4).
"almost randomly", I got 17784 minimal puzzles, including the automorphic one

Did you also have a "self trump" ?

It seems likely that you had the (one) puzzle I'm refering to.
It's one of those "SE anomalies".

There were 142 of those, which adds another 149 (puzzle, rating) pairs to list, for a total of 20871 "essentially distinct" pairs.

let's start a parallel thread just to collect SE anomalies

Must be why I said was calling them "SE anomalies" ... ... interesting.

Would you like to start one ? Would you like me to ?
I found 3 places where "dup-puzzle" penalties were incurred in the "2.0" game, where the "dup" had a defferent rating than the original -- the one in this game, and then here and here.

Edit: If one is started, I'ld rather see you do it, since you know more about SE, and the patterns game is "your turf".

[champagne]

Did you also have a "self trump" ?

It seems likely that you had the (one) puzzle I'm refering to.
It's one of those "SE anomalies".

I don't know how JPF is working, but I would expect that, as me, he first eliminates equivalent puzzles.

In fact, I have no chance to detect a "self trumping anomaly in SE rating".

The anomaly appeared here just due to the fact that the first puzzle was posted by Mike.

champagne

[ano1]

I don't know how JPF is working, but I would expect that, as me, he first eliminates equivalent puzzles.

I expect that's true.

Whether he has it, would depend on which canonical form he uses to store the "unique" puzzles.

[ Edit: I should correct that statement. If he checks for a duplicate using some canonical form,
but rates the puzzle in the form it was produced in, then the canonical form he uses to track
the unique puzzles, would be irrelevant. That's true too, for Mike's 9.3/9.3/3.4, below. ]

I collected the minlex versions of the 4 (in this case) "same shape, geometric transformations" of the puzzles.
That version of this one puzzle, rates as a trump.

For the 9.3/9.3/3.4 puzzle that Mike posted, the "minlex" form that I initially had, gave the 9.2/9.2/3.4 rating.
(Evidently he's using some other canonical form).

Glen may know more about this, but it's possible, too, that different digit assignments may lead to different ratings.
A quick and easy way to improve ratings consistency, though, would be to minlex-renumber the puzzles "first".
It's possible that SE is already doing that, or doing something equivalent.

[gsf]

Glenn may know more about this, but it's possible, too, that different digit assignments may lead to different ratings.
A quick and easy way to improve ratings consistency, though, would be to minlex-renumber the puzzles "first".
It's possible that SE is already doing that, or doing something equivalent.

I know enough about SE to have added the options for the patterns game
on my todo list (but at the bottom) is a wish to translate the dog slow java implementation to C
and then put some more minds on the task of fixing some of its shortcomings (like k-loops in the hardest puzzles)

I'm pretty sure it does not canonicalize before rating
that would be a reasonable way to fix the anomalies
but I wouldn't do that to the java implementation
its already the hot spot in probably everyone's patterns game arsenal

feel free to start a SE rating anomaly thread under interactive games

[champagne]

[ Edit: I should correct that statement. If he checks for a duplicate using some canonical form,
but rates the puzzle in the form it was produced in, then the canonical form he uses to track
the unique puzzles, would be irrelevant. That's true too, for Mike's 9.3/9.3/3.4, below. ]

This is for sure the right sentence;

Having to submit the original puzzle, rating a canonical form would create new problems.

SE RATING could work on a canonical form, but this is not at all in the line of a solver.

Anyway, the main problem with SE Rating is this endless computation as soon as you pass 9.5.

In the range of 10.5 to 11.5, my solver qualify a puzzle in 10 to 30 seconds, SE in 20 to 50 minutes(may be more). With patterns producing many high rating, filtering generated puzzles is a key issue.

champagne

[JPF]

::: comment :::

ano1,
For a new found puzzle:
I (normally) check there is no isomorphic puzzle in my list and in g.r.emlin's list.
Then, I store it and rate it (if necessary) with SE on a short minlex form.

short minlex form means :
If P is a puzzle P = a1 a2 ... a81, the short minlex form is a remapping of the digits into Q = b1 b2 ... b81 such that Q is minimum.

Now, about the game and Sudoku Explainer :
One of the key rule of the game is "to submit a new puzzle not isomorphic to any previous puzzle in the current game".
It's the player's duty to check it before submitting a puzzle.

The game is based on the SE ratings.
We all know the S. Explainer limits*, but it's by far the best publicly avalaible solver able to rate any puzzle and to give a solution.

feel free to start a SE rating anomaly thread under interactive games

I agree.

JPF

*see here

[m_b_metcalf]

Mike, I liked your X's ... here and on the 18's thread. I had never tried solving a 9x9 X before. I got the easy ones, but got stumped early on the hard ones. I'm not much of puzzle solver. Please tell me they needed something beyond locked sets -- finless fish at least.

Sorry for the late reply: I've been away for a few days. I don't really know how to rate x-sudokus once all the basic techiques are exhausted. Those hard ones require, in my code, brute force to solve. However, I do seem to remember that gsf makes a more sophisticated analysis, so maybe he could oblige. One of our local newspapers, the "Berliner Zeitung", publishes one every day

Regards,

Mike Metcalf

[gsf]

::: comment :::

Those hard ones require, in my code, brute force to solve. However, I do seem to remember that gsf makes a more sophisticated analysis, so maybe he could oblige.

your memory is better than mine
I have 2 solver variants that can do X
can you post a few dozen or so to test
thanks