Sodyoutoo (a.k.a. Sudoku)

[petrw1]

Originally Posted by biwema

If someone would like to send in the example, feel free to do so.

Done...

[petrw1]

Originally Posted by biwema

Some Sudoku can be quite tricky. The following example has exactly one solution, but i am not sure if there is a solving strategy (except brute force).
Even the sudoku solver (mentioned above) failed.

12.34.5..
3....6...
.54....3.
24....3.7
..5.....2
...428...
4..6..72.
.1......6
...9.....

Does someone know a strategy, that can solve this one?

http://www.sudokuwiki.org/sudoku.htm...6....31....4.7......

This solved it but had to use a lot of advanced steps and in the end had to resort to recursive trial and error to determine it in fact does only have 1 solution.

[Batalov]

Yes, this one is truly a monster. But it is very good that AI progress is evident, as time goes by -- and that the time capsule of this thread shows how something that was not possible for AI now became possible.

Similar progress in Go AIs is simply breath-taking. I clearly remember times when 5 kyu level for computer players on KGS was thought of as a plateau and that the technology was exhausted. Now they went over 1 dan (and I haven't even checked back for a year; maybe more now).

[Batalov]

Originally Posted by biwema

Some Sudoku can be quite tricky. The following example has exactly one solution, but i am not sure if there is a solving strategy (except brute force).
Even the sudoku solver (mentioned above) failed.

12.34.5..
3....6...
.54....3.
24....3.7
..5.....2
...428...
4..6..72.
.1......6
...9.....

Does someone know a strategy, that can solve this one?

I solved it with coloring. 'Enjoy sudoku' App really saves a lot of paper. After examining the pencil'ed state, I chose one entrance into the maze (it was 3 in c4 square), and then if it would have not solved it, I would have backtracked but I didn't need to; it basically feel through almost to the end.
Now, the question is - how do you choose an 'achilles heel'? You choose something that has a potential to decrease entropy significantly (you can see that from the first consequents). That's what I think at least.

[LaurV]

Grrr... I am still struggling with this one, being in a wall right now. I met the first wall last evening about half hour of playing with it "logically". I decided to do trials. In that point I knew already that the third cell in the first row is either 6 or 7, so I tried 6. It was still going for a while and so I switched the paper and tried 7. Picked them systematically, in order Fortunately this locked fast with an impossibility, so I know that the cell is 6. But I am in a place where there I see no chain, I can make no inference and I am sure there is no one short enough to be seen. I put about 30 minutes more into it this morning, for a total of about 2 hours (edit: more like 1 hour and 40 minutes, or so), but futile to say that this morning I couldn't make any progress.