Author |
Message |
pnm
Posted on: Tue Sep 11, 2012 8:37 pm
Posts: 3301 Joined: Thu May 12, 2011 11:58 pm
|
solver needs work...
It's obvious my killer sudoku solver can be improved a lot...: this puzzle ("Daily No. 2456"): http://killersudokuonline.com/play.html ... &year=2012is qualified as "Hard" on that site, but as too difficult by my solver (it does solve it, but it takes too much time)
|
|
|
|
pnm
Posted on: Tue Sep 11, 2012 10:39 pm
Posts: 3301 Joined: Thu May 12, 2011 11:58 pm
|
Re: solver needs work...
Do people have an opinion on that puzzle? I ran it through this, quite sophisticated, solver: http://www.sudokuwiki.org/killersudoku.htm, but after 30 steps or so it says "Run out of known strategies"...
|
|
|
|
rossiniman
Posted on: Wed Sep 12, 2012 3:41 am
Posts: 43 Location: Tacoma, WA, USA Joined: Fri May 13, 2011 4:03 am
|
Re: solver needs work...
Is there a way to print that puzzle?
|
|
|
|
clm
Posted on: Wed Sep 12, 2012 8:40 am
Posts: 856 Joined: Fri May 13, 2011 6:51 pm
|
Re: solver needs work...
pnm wrote: Do people have an opinion on that puzzle? I ran it through this, quite sophisticated, solver: http://www.sudokuwiki.org/killersudoku.htm, but after 30 steps or so it says "Run out of known strategies"... I solved the puzzle using the usual ways (innies and outies, etc., if you need the solution pse tell me), it's difficult but doable; i5 = 3, a4-b4 = [5,9], a6-b6 = [7,8] are inmediate; the first "unknown" number to place is e6 = 9: two possibilities, 6 or 9, considering that cage 30 must have 6,7,8,9 and that [7,8] are already in a6-b6; since e6 + h6 + i6 = 18 (outies), e6 = 6 >>> h6 + i6 = 12 with h6 = 3, i6 = 9 would be impossible due to a 9 in cage 29. Etc., it requires some time, small equations and analysis as usual but I think the level of difficulty is about the same as some of your book's puzzles. I can confirm the solution is unique.
|
|
|
|
bram
Posted on: Sat Sep 15, 2012 5:53 am
Posts: 253 Joined: Tue May 24, 2011 4:55 pm
|
Re: solver needs work...
I managed the "Daily No. 2456" at killersudokuonline.com with some difficulty. Even if the indicated difficulty level is only "hard" (not "mind-bending" or "outrageous") I found it clearly harder than the difficult killer sudokus here at calcudoku.org, which I am able to solve on-screen, memorizing candidate numbers when necessary and doing the calculations in my head. The "Daily No. 2456" I wouldn't have managed without using the option for entering candidate numbers on-screen because I found it too hard to keep all those intricate dependencies in my head.
It makes sense that the difficult puzzles in the last part of your new killer sudoku book would be at a "harder-than-difficult" level similar to that of the "Daily No. 2456" (as noted in clm's comment) because they are meant for being solved using a print (or the paperback) so puzzlers can enter candidate numbers in cells and make notes and calculations in the margins. The charming thing about book puzzles is the "armchair" feeling of solving them away from the computer and immersing yourself in pen-and-paper analysis.
|
|
|
|
pnm
Posted on: Sat Sep 15, 2012 10:37 am
Posts: 3301 Joined: Thu May 12, 2011 11:58 pm
|
Re: solver needs work...
bram wrote: I managed the "Daily No. 2456" at killersudokuonline.com with some difficulty. Even if the indicated difficulty level is only "hard" (not "mind-bending" or "outrageous") I found it clearly harder than the difficult killer sudokus here at calcudoku.org Yes, so far the difficult killer sudokus on the site are clearly easier than the difficult ones of the book. With a few more solver improvements I managed to get the solving time down by a factor of 20 Two more obvious improvements to go, after which I should be able to rate this puzzle better. and thanks for the feedback :)
|
|
|
|
pnm
Posted on: Sun Sep 16, 2012 5:02 pm
Posts: 3301 Joined: Thu May 12, 2011 11:58 pm
|
Re: solver needs work...
pnm wrote: With a few more solver improvements I managed to get the solving time down by a factor of 20 And after a few more (quite fundamental) updates, the solver runs another 30 times faster. So in the end solving time went down from ~ 10 minutes to less than 1 second (!) Now I need quite a fundamental update to the way I determine a puzzle's difficulty as well...
|
|
|
|
pharosian
Posted on: Sun Sep 16, 2012 7:13 pm
Posts: 116 Joined: Sat May 14, 2011 3:18 am
|
Re: solver needs work...
pnm wrote: So in the end solving time went down from ~ 10 minutes to less than 1 second (!)
Congratulations! Well done! pnm wrote: Now I need quite a fundamental update to the way I determine a puzzle's difficulty as well... Yes... I just went through the calendar, charting number of times solved against difficulty level for the Difficult Killer Sudoku. The scatter plot shows no correlation between the two. I realize other factors (such as day of week) may be in play, but with a more accurate difficulty rating, I would expect to see more times solved for easier puzzles and fewer times solved for harder puzzles.
|
|
|
|
|