User-friendly internet only solver at sudokusolver.co.uk

dagojr · Posted: Fri Jan 26, 2007 5:56 pm Post subject: Logical solution to #27

I was looking at "illogical" puzzle #27 on Sudoku Solver's website. These are sudoku that the online solver cannot solve with its current code without guessing. However, I think I've found a simple and logical way to finish this one.

Here is the original puzzle:

coloin · Posted: Sun Jan 28, 2007 12:09 am Post subject:

Well I dont go in for solving but with the help of sudoku explainer 1.2 this is what it said !

The notation is the opoosite to yours !

Phoxtrot · Posted: Wed Jul 11, 2007 11:32 am Post subject: Illogical #13

Hello,
I have had about the same idea as you and made a logical sudoku solver too Smile

I checked my program against the 18 first illogical sudokus you posted.

I deactivated my postulate and my guess algorithms for that purpose.
(I make a distinction between the 2: In postulate, i try a value in a square but the only situation where I consider I have learned something is if it leads me to an invalid grid. In guess algorithm, I do kind of a depth first search, trying values along the way until I get a solution.)

From the 18 first sudokus, I was only able to solve the 13th under those conditions. Only 2 non trivial algorithms were used:

TupleGreed5:

Screenshot of the grid at before algo is used:
(Not allowed to post links, so skipping around the limitation:the screenshot is on imageshack, tag sudoku)

DEBUG - (su1.algorithm.TupleGreed) Tuple(5): [1,8][1,7][1,6][1,4][1,1] union:2 5 6 8 9

In the 5 squares noted above ([row,column], top left is [0,0]), only the values 25689 are allowed; since the are 5 such squares in that line, the other squares in that line cannot have any of those values.

Note: there is a dual algorithm: You could also say that value 347 are only allowed to 3 squares on that lines. Which implies that those 3 squares must have one of these values. I didn't program that algorithm since it will solve the same situations as its dual... (with a different number of involved squares)

BlockToCol:

You probably already have that algorithm so I won't go to the trouble of making a screenshot unless necessary.

DEBUG - (su1.algorithm.BlockToRowCol) BlockToCol around [1,1] for value:5

Meaning that for the block around position [1,1] (that is the upper left block of 9 squares), the value 5 can only be found in column 1 (again counting from 0). Therefore, value 5 is forbidden to the squares in that column that do not belong to the upper left block.

If you have any question, feel free to ask Smile

I have a few tough algorithms in my program. (Though i programmed those some time ago and i might have trouble remembering exactly whta they did...)

Phoxtrot · Posted: Wed Jul 11, 2007 2:53 pm Post subject:

Ohoh, I have added the XY wing algorithm explained above and now i can solve the illogical #16 too.

The only 2 difficult algorithms it used are the XY-wing alogrithm (which i renamed crossSection3_0, go figure why...)

DEBUG - (su1.algorithm.CrossSection3_0) CrossSection3 (row) info on square [8,6] for value:3 by crossSecting (size2) 3 squares: [6,1];[8,2]&[6,6]

And a BlockToCol algorithm (as in my previous post but for a column instead of a row)

sardonic · Posted: Thu Aug 30, 2007 1:21 pm Post subject:

I dont know bout you guys, but I dont solve sudoku's by first entering all numbers in all open squares and then remove the numbers allready taken. It would take ages this way and I'm pretty sure, you only use this method once you get stuck using the 'normal' method, like most of us do.

So I think most of us solve them the exact opposite way:

64__139__+1___264__+_29_457__+__2___83_+86__37_19+7__2_9___+__13__69_+9364_8_2_+__5______

First of all, I've given each big square (of 9 smaller squares) a number from 1 to 9, counting from left to right, top to bottom:
S1 S2 S3
S4 S5 S6
S7 S8 S9

Ok, lets start solving:

- S1 has a 1 at row B, S2 has a 1 at row A, so S3 should have a 1 in row C
- C7 is already taken and row 8 already has a 1, so C9 is the missing 1

- S1 has a 6 at row A, S2 has a 6 at row B, so S3 should have a 6 in row C
- only 1 spot left, so C8 is a 6

- S1 has a 2 at row C, S2 has a 2 at row B, so S3 should have a 2 in row A
- A7 is already taken and row 8 already has a 2, so A9 is a 2

- S1 has a 9 at row C, S3 has a 9 at row A, so S2 should have a 9 in row B
- only 1 spot left, so B4 is a 9

I call this method "comparing squares". The next method is "completing squares":

- S2 still needs a 7 and 8
- row C already has a 7, so C4 is an 8 and thus A4 is a 7

Which leads us to the next method "completing rows":

- row C still needs a 3, only 1 spot left, C1 is a 3

If you can solve the sudoku using only these 3 methods, then I categorize this as Easy.

I would really like to see you program a sudoku solver that uses this method, as its more realistic. Who on earth solves sudoku's by first filling all the blank spots with numbers 1 to 9 Confused

I know this is kinda like starting all over, so maybe program a second sudoku solver, using 'human' logic only Wink

Pat · Posted: Tue Sep 04, 2007 12:35 pm Post subject:

Miroslav Kuzela · Posted: Mon Nov 12, 2007 8:04 pm Post subject: Solve Method

My question:
What are differences between the solve method C (difficulty score=16) and the solve method D (difficulty score=32) for value N=2 ???

rajkumar_innani · Posted: Fri Jun 05, 2009 5:28 am Post subject: SUDOKU SOFTWARE

I've developed tow softwares - one can solve the given puzzle and the other can generate millions and millions of new puzzles with their solutions. For further information contact direct No. +91 9448588369

hendrik

Hello Developers,

i hacked a little sudoku solver by myself for fun and now found your sudoku solver.
Your code solves the following sudoku that my app cannot solve:

000000018700040000000000030420000700000001000000300000500070200601800000040000000

I checked your solve log and in found the place where the magic or the
bug happened (see below). Can you explain what happened there? I came
to the same upper block and can not see how to go further with solving.

Thank you and greetings
Hendrik

m_b_metcalf · Posted: Fri Apr 16, 2010 12:19 pm Post subject:

hendrik · Posted: Fri Apr 16, 2010 1:28 pm Post subject:

Hello Mike,

I understand how to go on with the second table. But i did not see how it is possible to come from the first table to the second table. I think there is maybe a bug in the solving code.

Greets,
Hendrik

m_b_metcalf · Posted: Fri Apr 16, 2010 1:43 pm Post subject:

I think not. For instance:

hendrik · Posted: Fri Apr 16, 2010 1:53 pm Post subject:

Sure, you are right! Now i understand this. thank you!

Greets,
Hendrik