The New Sudoku Players' Forum

[JPF]

For sure, I know these commands but there is something strange in the program :

Take this champagne’s puzzle from the last game :

Code: Select all

001050200030004010700000304000600040300080100010009000502070001060900050003000800

Rating =10.5/10.5/2.6
If I use the command line --diamond , it takes almost 220 seconds to get the answer 20.0/20.0/2.6

After the first rating (ED=2.6), the only thing the program has to do is to get the rating of the next candidate elimination, which is 9.9
It takes 1 second.
After that, the program can return 20.0/20.0/2.6

JPF

[gsf]

For sure, I know these commands but there is something strange in the program :

Take this champagne’s puzzle from the last game :

Code: Select all

001050200030004010700000304000600040300080100010009000502070001060900050003000800

Rating =10.5/10.5/2.6
If I use the command line --diamond , it takes almost 220 seconds to get the answer 20.0/20.0/2.6

After the first rating (ED=2.6), the only thing the program has to do is to get the rating of the next candidate elimination, which is 9.9
It takes 1 second.
After that, the program can return 20.0/20.0/2.6

thanks for the counter example
that will make it easy(ier) to track down
time to add some regression tests ...

[JPF]

I don't know what you mean.

With the --diamond option, the actual program calculates ER in order to check if ER=ED.
Again this is a waste of time as soon as a candidate elimination (before the first placement) requires a technique rated higher than ED.

Is it really a huge task to adjust the program for that or even better to return only ED?

JPF

[gsf]

I don't know what you mean.

With the --diamond option, the actual program calculates ER in order to check if ER=ED.
Again this is a waste of time as soon as a candidate elimination (before the first placement) requires a technique rated higher than ED.

Is it really a huge task to adjust the program for that or even better to return only ED?

I meant:
there's a bug
your example exposed the bug
having an example to work off of makes it easier to find and fix the bug
retaining the example as a test will guard against a regression to the same bug after future modifications
thanks

[JPF]

Ok;it's the words "regression tests" I didn't understand.
Thanks.
JPF

[gsf]

Ok;it's the words "regression tests" I didn't understand.
Thanks.
JPF

aha
one of the more important things for a long living piece of software is a good
collection of regression tests -- after a certain code size changes to the code
are rarely independent

[999_Springs]

"cat | sed | sort | uniq | sort" of the results for the first 114 games yields these counts of accepted submissions. Only eleven players have reached double-digits.

Code: Select all

   3624 gsf
   2191 m_b_metcalf
   1909 JPF
    809 TTHsieh
    800 Mauricio
    655 champagne
    525 joel64
    313 tarek
    256 Pat
     51 coloin
     31 999_Springs
      9 ronk
      9 mario133
      5 Ruud
      5 Eioru
      4 wintder
      2 g.r.emlin
      1 pseudocoup
      1 draco
      1 claudiarabia

... Wow. 31!? I didn't know I'd wasted so much time randomly clicking on my copy of SE that wasn't even fully responsive to mouse clicks.

[ronk]

Some comments on this pattern.
I got only 6983 different puzzles.

Dang, I must have done something seriously wrong. I did a random search for the entire game time + 24 hours after the last submission ... and found only 536 ... using this template at 2.8 GHz.

Code: Select all

 1        .        .        | .        2        .        | .        .        3
 .        234      .        | .        .        .        | .        12456    .
 .        .        2345     | .        .        .        | 124567   .        .
----------------------------+----------------------------+----------------------------
 .        .        .        | X        .        X        | .        .        .
 2345678  .        .        | .        X        .        | .        .        X
 .        .        .        | X        .        .        | .        .        .
----------------------------+----------------------------+----------------------------
 .        .        X        | .        .        .        | X        .        .
 .        X        .        | .        .        .        | .        X        .
 X        .        .        | .        X        .        | .        .        X

Were my clue restrictions faulty?

[JPF]

6 is missing in r3c3

JPF

[ronk]

6 is missing in r3c3

So moving the 6 from r3c3 to r2c8 is invalid? I don't understand why.

[JPF]

I don't know
Here are some puzzles with r3c3=6

Code: Select all

100020003040000050006000100000306000200010007000500000001000800050000060300090002
100020003040000050006000100000306000200070008000500000002000700050000060300090002
100020003040000050006000100000306000200070008000500000008000700050000060300090002
100020003040000050006000100000401000200070008000600000005000400090000060800030001
100020003040000050006000100000402000700030006000500000008000400050000090600070001
100020003040000050006000100000405000300060007000700000008000400050000060700030002
100020003040000050006000100000405000300060007000800000008000400050000060700030002
100020003040000050006000100000405000300070002000600000007000400050000080200080009
100020003040000050006000100000405000300070006000600000008000400050000070600030002
100020003040000050006000100000405000300070008000600000007000400050000090800030001


JPF

[ronk]

Sorry, I forgot a filter was discarding puzzles with difficulty rating less than 200. I'll have to check that against the ratings of accepted submissions.

[champagne]

I don't know why JPF did not play that one...

This puzzle popped up after g.r.emlin's deadline.
I started the game too late; it takes time for my programs to be really efficient.
By the way, the rythm of the game is a serious hurdle for potential newcomers.

JPF

I think this is the right place to comment on that post.

When I entered the game some months ago, I already found the entry barrier terribly high. To-day, it's worse.


A newcomer has a lot to do to enter the game.

a) build an efficient organization to handles huge files
b) find an efficient process to generate puzzles, focusing on most promising areas
c) extract puzzles to be rated by SE to turn around that bottleneck.

......

a) is specific to each of us.

A significant part in 'b' and 'c' is influenced by the "strategy" followed by the players (trying to figure out what the pattern will produce), but most of the performance is done by the specific tools each of us has implemented.

The only way to lower the barrier for newcomers is somehow to share the processes and the code we are using.
I have personnaly no problem to go in that direction.

The main drawback in my case is that I have only one program doing all tasks, not stable at all.
This is not so easy to share unless somebody is prepared to work hard on the code.

I thought processes would be easier to share but i opened the door recently and no new question came;

champagne

[gsf]

I started the game too late; it takes time for my programs to be really efficient.
By the way, the rythm of the game is a serious hurdle for potential newcomers.

I still use my solver as described in the first post in this thread
there's some unix shell script glue that passes generated puzzles to sudoku explainer
plus another shell script that makes a game submission

re: the game rythym
the first post in the games thread has the game parameters
can we come up with a newcomer set of game parameters and use that every so often (like lightning vs. 10 min between moves)?
would increasing the time between moves (WAIT_SUBMIT) be sufficient to even the playing field?

[champagne]

To continue to open the door on how I work, some statistics on game 117.

I generated puzzles exclusively using a -n +n process and I stopped generation after about 30 hours.
At that moment, I had still unrated puzzles with an ED around 2.0;2.6.
As JPF had found the highest rating with such an ED, I rated that lot to close the game.
As I have for the time being no quick filter for High ER and ED=1.2, I did not rate these puzzles.

In fact, after about 15 hours, i focused generation on puzzles using at least a pair for the first elimination.
(this is easy to do in a -n+n process).

I generated 445 246 different puzzles.
I rated only 6453 puzzles having an ER of 8.0 and more

out of these puzzles, some interesting stats on the situation at the end of the game

Code: Select all

er    ep   ed    number in cache

9.6   9.6  2.6   3

9.5   9.5  3.0   1
9.5   9.5   2.6  8

9.4   9.4  2.6   28

9.3   9.3   3.4   1
9.3   9.3   2.6 138

9.2   9.2  3.4   1
9.2   9.2  3.0   4
9.2   9.2  2.6 201

9.1   9.1  3.8    3
9.1   9.1  3.0    5
9.1   9.1  2.6 363

9.0   9.0  3.8    2
9.0   9.0  3.4    6
9.0   9.0  3.0   9
9.0   9.0  2.6 687

champagne