Swordfish question

Advanced methods and approaches for solving Sudoku puzzles

Swordfish question

Postby datprogrammer » Wed Jun 22, 2005 8:50 pm

I'm very interesting in simes' solving techniques but I am puzzled by the swordfish.., in the example quoted, he states:

"Look for three columns with only two candidates for a given digit.." and then

"At this point, we have a swordfish in the 1s. There are two candidate positions in column 6 - (2, 6) and (6, 6). "

I've studied and studied everywhich way but I don't get this... surely, in the highlighted columns, the cell (1,1) is a candiate for a 1, and in column 4, we have (2,4), (3,4) and (8,4) also as candidates for a 1 and column 6 has additional candidaes for a 1 in (3,6) and (8,6)

I'm missing something obvious but I can't see it! Help!
datprogrammer
 
Posts: 18
Joined: 22 June 2005

Postby IQ » Wed Jun 22, 2005 9:10 pm

I haven't actually had a look at simes' solving techniques page yet but I think it has something to do with the 1 corresponding with a 1 in the same column above or maybe below it?
IQ
 
Posts: 6
Joined: 27 May 2005

Re: Swordfish question

Postby scrose » Wed Jun 22, 2005 9:11 pm

datprogrammer wrote:I'm missing something obvious but I can't see it! Help!

Look at where the candidates 3 and 5 can go in row 1.
Look at where the candidates 1 and 8 can go in box 2.
Look at where the candidates 1 and 6 can go in box 8.
scrose
 
Posts: 322
Joined: 31 May 2005

Postby simes » Wed Jun 22, 2005 9:50 pm

swordfish is a more complex x-wing, so ensure you understand x-wing before reading about swordfish.

Swordfish is about linked cells. It occurs when you have a pattern of cells that allow you to say "If cell x is N, then cell y cannot be N, so cell z must be N" etc. Forming a chain of interconnected cells. Since cells x, y and z share columns and rows, you can eliminate candidates for other cells in those rows.

(That's probably not much clearer than the original.)
Last edited by simes on Sun Dec 11, 2011 9:53 am, edited 1 time in total.
simes
 
Posts: 324
Joined: 11 March 2005
Location: UK

OK but...

Postby datprogrammer » Wed Jun 22, 2005 11:46 pm

<Editing after considering Scrose's comments>

I understand the x-wing, what I'm not getting is how the 1's can be eliminated from the cells mentioned in my original question.

Simes, so your example is unclear. It's not immediately obvious that 1 cannot be placed in (1,1) - until you realise that there's a disjoint subset of (3,5) in (1,1) and (1,3) (Thanks Scrose!) - perhaps you should point that out in your example?

I don't follow Scrose's 2nd and 3rd hints -

"Look at where the candidates 1 and 8 can go in box 2. "
8 - (2,5) (2,6), (3,5),(3,6)
1 - (2,4),(2,5),(2,6),(3,4),(3,5),(3,6)
How does that help?!

"Look at where the candidates 1 and 6 can go in box 8."
1 - (8,4),(8,5),(8,6),(9,4)
6 - (8,4),(8,5),(9,4)
Again, how does that help?

Steve (Still confused!)
datprogrammer
 
Posts: 18
Joined: 22 June 2005

Postby scrose » Thu Jun 23, 2005 2:14 am

The candidates for row 1 are as follows.

{1358} {18} {35} [4] [7] [9] [6] [2] {18}

Notice that the candidates 1 and 8 can be found as a pair in the cells r1c2 and r1c9. Thus, the candidates 1 and 8 can be removed from the other cells in row 1. The remaining candidates for row 1 are as follows.

{35} {18} {35} [4] [7] [9] [6] [2] {18}

The candidates for box 2 are as follows.

[4] [7] [9]
{126} {1268} {18}
{135} {18} {1358}

Notice that the candidates 1 and 8 can be found as a pair in the cells r2c6 and r3c5. Thus, the candidates 1 and 8 can be removed from the other cells in box 2. The remaining candidates for box 2 are as follows.

[4] [7] [9]
{26} {26} {18}
{35} {18} {35}

The candidates for box 8 are as follows.

[8] [9] [4]
{1367} {16} {137}
{16} [5] [2]

Notice that the candidates 1 and 6 can be found as a pair in the cells r8c5 and r9c4. Thus, the candidates 1 and 6 can be removed from the other cells in box 8. The remaining candidates for box 8 are as follows.

[8] [9] [4]
{37} {16} {37}
{16} [5] [2]

Enough candidate 1's have now been eliminated to permit the swordfish technique to be applied.
scrose
 
Posts: 322
Joined: 31 May 2005

Postby datprogrammer » Thu Jun 23, 2005 6:32 am

Scrose,

Thank you! I did what I should have done in the first place, - dubbed it into wayne's program and pencil marked it - it then becomes obvious.

My only (pathetic) excuse is it was very late and I had been on a mind-numbing training course at work all day (function point counting - dull!)

I think the description associated with the example on simes' page should point out that the 1's can be eliminated from the other cells by using the pairs technique - just a suggestion for Simon if he is reading!

One final question for anyone reading this, how do I get Wayne's program to generate "very hard" puzzles? When I dubbed this one in and validated it, it rated as "very hard" - simes' notes state that the program does not use sworfish in it's bag of generation tricks, but that x-wing IS used - the program also grades the x-wing examples as "very hard" - so how do I generate such puzzles?
datprogrammer
 
Posts: 18
Joined: 22 June 2005

Postby simes » Thu Jun 23, 2005 8:53 am

...just a suggestion for Simon if he is reading!

I'm reading, and will update that page.

...simes' notes state that the program does not use swordfish in it's bag of generation tricks

More clarificaion required I think. My note is meant to refer to my OWN program, not Pappocoms!
Last edited by simes on Sun Dec 11, 2011 9:53 am, edited 1 time in total.
simes
 
Posts: 324
Joined: 11 March 2005
Location: UK

Postby scrose » Thu Jun 23, 2005 1:36 pm

datprogrammer wrote:how do I get Wayne's program to generate "very hard" puzzles?

First, either press F8 or activate the Hard button on the toolbar. Then, while holding down Ctrl, either press F5 or click the New button on the toolbar.

You can find more information about special grades by accessing the help by pressing F1. From the Contents, go to The details > Grades of puzzles > Special grades.
scrose
 
Posts: 322
Joined: 31 May 2005

Thankyou all!

Postby datprogrammer » Thu Jun 23, 2005 2:48 pm

simes, sorry about the confusion! I assume your program switches to brute force mode then when the only logical way forward is to spot the swordfish?

Does anyone know if Wayne's program generates puzzles requiring swordfish?

One last question - I was just wondering if there are any "ballpark" figures available for soving times - I've seen a couple of websites where they record your solving time and show a league table but that is cleary abused with the use of automation. I'd like to know how I'm shaping up in terms of solving times using Wayne's graded puzzles as a starting point.

oh, and is there a 12 step program for coming off Suduko addiction ?! ("Hello, I'm Steve and I'm a Sudoku-holic!")
datprogrammer
 
Posts: 18
Joined: 22 June 2005

Postby Animator » Thu Jun 23, 2005 2:54 pm

Pappocom's program will not generate a puzzle that requires the Swordfish technique.

In some puzzles however you can spot a swordfish... but you don't need to see it to remove a candidate from a cell. Or to be correct, it will not be the only way to remove the same candidate from a particular cell. An easier rule will cover it too.

Also note that the question about why you can elimnate some candidates from that example is asked before...

Update, a discuss in which it was discussed as well (or atleast partially) http://forum.enjoysudoku.com/viewtopic.php?p=2219&highlight=#2219
Animator
 
Posts: 469
Joined: 08 April 2005

Average Times and Su Dokuholics Anonymous

Postby george-no1 » Thu Jun 23, 2005 5:29 pm

datprogrammer wrote:I was just wondering if there are any "ballpark" figures available for solving times?


I haven't seen any figures quoted on this site, but I (try to) do the Times Su Doku as following:

Easy: Under 4 minutes
Mild: Under 9 minutes
Difficult: Under 15 minutes
Fiendish: Under 23 minutes

Those are are about average times for me, but I'm sure that there are many people out there who can do a Fiendish in under 10 minutes. I am definitely not one of these people, although I am a great believer that practice makes perfect, and the more you do, the better you will become. For instance, about a month ago I could not do a Fiendish in less than 40 minutes and now my time is less solely because I've become hooked and I do 1 or 2 every day without fail.

As for your second question:

datprogrammer wrote:oh, and is there a 12 step program for coming off Suduko addiction ?! ("Hello, I'm Steve and I'm a Sudoku-holic!)"


So far I have not heard of one, but I'm sure, given time, they will develop all over the country.

George:)
george-no1
 
Posts: 150
Joined: 20 May 2005

Postby Pappocom » Thu Jun 23, 2005 5:47 pm

datprogrammer wrote:One last question - I was just wondering if there are any "ballpark" figures available for soving times - I've seen a couple of websites where they record your solving time and show a league table but that is cleary abused with the use of automation. I'd like to know how I'm shaping up in terms of solving times using Wayne's graded puzzles as a starting point.

You could do the Time Trial on this site and see how your times compare with others'. It's in the Library books and puzzles Forum.

- Wayne
Pappocom
 
Posts: 599
Joined: 05 March 2005

Postby scrose » Thu Jun 23, 2005 5:56 pm

datprogrammer wrote:I'd like to know how I'm shaping up in terms of solving times using Wayne's graded puzzles as a starting point.

The problem with trying to compare solving times amongst players is agreeing on a set of solving techniques available to all players. The "Hall of Fame" in the Pappocom software attempts to tackle this problem by only letting you enter a time into the hall of fame if you are playing by specific rules. So after you've got five or so times listed in your hall of fame, you will start to get an idea of your own average time for a given difficulty grade.
scrose
 
Posts: 322
Joined: 31 May 2005

Postby datprogrammer » Thu Jun 23, 2005 11:25 pm

"The problem with trying to compare solving times amongst players is agreeing on a set of solving techniques available to all players"

Why is that a problem - surely the techniques don't matter - the challenge is to "solve this puzzle as fast as possible" Obviously we can assume we mean that no solving software is to be used, i.e just brain power, but other than that, what does it matter if I use one set of techniques and you use another - we're just comparing who can solve fastest...
datprogrammer
 
Posts: 18
Joined: 22 June 2005

Next

Return to Advanced solving techniques