Ulterior Puzzles

Everything about Sudoku that doesn't fit in one of the other sections

Postby tarek » Wed May 31, 2006 7:33 pm

thanx Ruud, I fixed a number of bugs, The Log that you provided coupled with the ratings of RW's #3 & #4 puzzles were helpful.

I've now replicated your results for all RWs puzzles.....

I have added the Fruitfulness formula to the solver so RW's puzzles have the following
Code: Select all
       Steps  B1    Fruitful Ratio
#1     15     5     2        0.4
#2     17     6     2        0.333
#3     12     11    6        0.545
#4     15     8     3        0.375
#5     9      7     4        0.571
#6     10     6     4        0.666
where B1 is the number of steps where Locking candidates was possible.

Ocean's beautiful 27 Stepper & 25 Steppes
Code: Select all
       Steps  B1    Fruitful Ratio
       27     13    2        0.153
       25     8     4        0.5


& my latest 2 puzzles
Code: Select all
       Steps  B1    Fruitful Ratio
#1     19     9     5        0.555
#2     22     7     2        0.285


How about multiplying the ratio by the number of Steps to give us an overall score (ulterior motives:) )???

The clear winner until now would be Ocean's 25 Stepper Scoring 12.5 (RW that should make you happy)

tarek
User avatar
tarek
 
Posts: 2622
Joined: 05 January 2006

Postby RW » Thu Jun 01, 2006 9:44 am

tarek wrote:The clear winner until now would be Ocean's 25 Stepper Scoring 12.5 (RW that should make you happy)


Yes, that makes me very happy.:D

I'm not quite sure if the amount of fruitful pointing pairs should be compared to the amount of possible pointing pairs. Why should puzzles with more possible (non fruitful) pointing pairs be punished with less points? Looking at my #1 and Ocean's 27:

Code: Select all
        Steps  B1    Fruitful Ratio  Ulterior score
My #1   15     5     2        0.4    6
Ocean's 27     13    2        0.153  4,131


Doesn't seem right in any way if a 15 stepper with 2 fruitful pointing pairs can score more than a 27 stepper with 2... IMO it would be better to either compare them to the amount of total steps or not to compare them to anything at all (just add the amount to the steps, possibly multiplied with some constant that you find suitable).

PS. Not that I expect it to be a winner, but it would still be nice to see some ratings on my #7.

RW
RW
2010 Supporter
 
Posts: 1000
Joined: 16 March 2006

Postby tarek » Thu Jun 01, 2006 12:05 pm

This is unofficial (awaiting the Threadmaster to confirm), but your #7 RW scored a 14, StepCount:0604030305080401020406050302, B1 possible steps: 5; Fruitful B1 steps: 3........

I'm not sure if the fruitful steps have any role really, it would be nice to see puzzles with fruitful step counts closer to the total number of steps though......

As you head towards low stepper ulteriors that gap should be smaller, but at increased step count, it should get more difficult to have that small gap & hence IMO puzzles with more Fruitfuls steps should score more.

tarek
User avatar
tarek
 
Posts: 2622
Joined: 05 January 2006

Postby tarek » Sat Jun 03, 2006 11:16 am

Time for some low steppers.....

8 Steps
Fruitful B1 Steps 6
Total B1 Steps 7
Ratio .857
Not minimal
180 Symmetrically minimal

Code: Select all
 . . . | 5 . . | . 3 . 
 . . 1 | . . . | 9 . 8 
 . 9 . | 1 . . | . 5 . 
-------+-------+------
 4 . 7 | . 1 . | . . . 
 . . . | 6 9 5 | . . . 
 . . . | . 7 . | 1 . 2 
-------+-------+------
 . 4 . | . . 3 | . 9 . 
 5 . 8 | . . . | 6 . . 
 . 7 . | . . 1 | . . . 


tarek
User avatar
tarek
 
Posts: 2622
Joined: 05 January 2006

Postby gsf » Sat Jun 03, 2006 1:35 pm

Ruud, can you point to the definitive message in this thread that describes how ulterior steps are counted?
thanks
gsf
2014 Supporter
 
Posts: 7306
Joined: 21 September 2005
Location: NJ USA

Postby ab » Sat Jun 03, 2006 2:13 pm

gsf wrote:Ruud, can you point to the definitive message in this thread that describes how ulterior steps are counted?
thanks


Here it is:

Ruud wrote:
The majority of support goes to ab's method. Since this is a democratic thread, we'll let the majority decide before the majority changes its mind.

Here's the technique:

1. Find all locked pairs (type 1), but pretend they do not exist.
2. Make eliminations based on the locked pairs found in step 1.
3. Find all hidden singles, but pretend they do not exist.
4. Place the hidden singles found in step 3 and perform eliminations.
5. Increment step counter and repeat from step 1.

Conditions:

a. The puzzle must be completely solved using the above steps (duuh)
b. The FN test must produce an (invalid) result.

ab
 
Posts: 451
Joined: 06 September 2005

Postby gsf » Sat Jun 03, 2006 3:42 pm

ab wrote:Here it is:
1. Find all locked pairs (type 1), but pretend they do not exist.
2. Make eliminations based on the locked pairs found in step 1.
3. Find all hidden singles, but pretend they do not exist.
4. Place the hidden singles found in step 3 and perform eliminations.
5. Increment step counter and repeat from step 1.


thanks
now I need a clarification from the RW#4 trace
I get a discrepancy at step 9
step 8 gets the puzzle here
Code: Select all
  9    4    5  | 167  167  16  |  3    2    8
 67    2   67  |  8    9    3  |  4    5    1
  8    1    3  |  5    2    4  |  7    6    9
---------------+---------------+---------------
 36    7    9  |  4  1356  16  | 156   8    2
 346   8   46  |  9  1356   2  | 156  13    7
  1    5    2  | 36    8    7  |  9   34   46
---------------+---------------+---------------
 47    3   147 |  2   146   5  |  8    9   46
  2    6    8  | 137 1347   9  |  1   14    5
  5    9   14  | 16   146   8  |  2    7    3

and the step 9 log is
Code: Select all
Step 9

Digit 1 locked in B9R8
Eliminate 1 in R8C4
Eliminate 1 in R8C5
Single in B9D6
Place 6 in R7C9

but it missed the lone 6 in box9 which can eliminate [69]^6
(all candidates (6) in one row/col within a box (9))
commit that move and it opens up a hidden single [64]=6 for step 9
gsf
2014 Supporter
 
Posts: 7306
Joined: 21 September 2005
Location: NJ USA

Postby ab » Sat Jun 03, 2006 5:47 pm

gsf wrote:but it missed the lone 6 in box9 which can eliminate [69]^6
(all candidates (6) in one row/col within a box (9))
commit that move and it opens up a hidden single [64]=6 for step 9

Personally I'd agree with you here, but it depends on how people have programmed their algorithm for looking for locked candidates. If they've assumed they'd need to find 2 or 3 candidates in a line then they'd miss this. In the normal course of solving puzzles this situation wouldn't arise, because you'd place the 6. Maybe it explains why I've been unable to match Ruud's step counts so far:!:
ab
 
Posts: 451
Joined: 06 September 2005

Postby gsf » Sat Jun 03, 2006 6:09 pm

ab wrote:Personally I'd agree with you here, but it depends on how people have programmed their algorithm for looking for locked candidates. If they've assumed they'd need to find 2 or 3 candidates in a line then they'd miss this. In the normal course of solving puzzles this situation wouldn't arise, because you'd place the 6. Maybe it explains why I've been unable to match Ruud's step counts so far:!:

time to do some 3rd party comparisons
are the current ulterior puzzles in one spot or spread out?
gsf
2014 Supporter
 
Posts: 7306
Joined: 21 September 2005
Location: NJ USA

Postby tarek » Sat Jun 03, 2006 10:31 pm

You are right gsf in saying that a box line elimination is possible. however, I actually wouldn't go for a box-line elimination if there is an obvious hidden single there.....I would place the hidden single only

[If two deductions can be made based on one cell then I would go for the more obvious or easier deduction first, if that was the hidden single then the step ends there. If it was the box-line elimination then you are correct in your assumption].

I think that the setting should be at >1 per line per box because I consider the hidden single EASIER than box-line elimination.

tarek
User avatar
tarek
 
Posts: 2622
Joined: 05 January 2006

re: RW ulterior puzzle #3

Postby Pat » Sun Jun 04, 2006 12:00 pm

RW (2006.May.29) wrote:[ 6 ulterior puzzles ]

#3
Code: Select all
 9 1 . | . 3 . | . 2 8
 4 . . | . . 7 | . . 3
 . . . | 8 . 5 | . . .
-------+-------+------
 . 4 3 | . . . | 8 . .
 6 . . | . . . | . . 5
 . . 9 | . . . | 4 7 .
-------+-------+------
 . . . | 2 . 9 | . . .
 1 . . | 3 . . | . . 4
 8 3 . | . 4 . | . 6 9


nice puzzle — does seem to need 4 exclusions from box to line.

the odd thing is, when i first solved it i was freely using both types of box-line interaction, as i usually do — and i found i had solved the puzzle with 3 exclusions from line to box — which is the type Ruud has dis-allowed for the present Topic — so i went looking for the equivalent exclusions of the other type, and i seem to need one more ( 4 ).


by the way, the Pappocom rating for puzzles #3 and #5 is Hard;
the other 4 puzzles are Medium.
Last edited by Pat on Sun Jun 04, 2006 8:25 am, edited 1 time in total.
User avatar
Pat
 
Posts: 3438
Joined: 18 July 2005

Postby tarek » Sun Jun 04, 2006 12:24 pm

You are solving it Pat in the method which this thread does not employ (we employ ab's method, Hidden & box-line at the same time).... (Pat's method, is exhausting hidden singles before going to type one box-line eliminations)......

The puzzle you mentioned solves in 12 steps(ab's method), 11 out of these steps you can make box-line eliminations out of which 6 should put more placements than if left for hidden singles alone......

tarek
User avatar
tarek
 
Posts: 2622
Joined: 05 January 2006

only 2 steps of "hidden singles"

Postby Pat » Sun Jun 04, 2006 1:44 pm

RW (2006.May.29) wrote:In this diagram 'X' represents any number except 'A':

Code: Select all
X X X|X . .|. . .
. X .|. . .|. . .
X . X|X . .|X . .
-----+-----+-----
. X X|. . .|X . X
. . .|. . .|X . .
. . .|. A .|. . .
-----+-----+-----
. X .|. . .|. . .
X X X|X X .|. . .
. . .|. . .|. A .


All digits 'A' can be solved without pointing pairs in 5 steps:

Code: Select all
X X X|X . .|5 . .
. X .|5 . .|. . .
X 4 X|X . .|X . .
-----+-----+-----
2 X X|. . .|X . X
. . .|. . .|X . 1
. . .|. A .|. . .
-----+-----+-----
. X 3|. . .|. . .
X X X|X X 4|. . .
. . .|. . .|. A .

- no, only 2 steps of "hidden singles"
  1. r4 r8 box6
  2. r1 r7 c2 c4
Code: Select all
X X X|X . .|2 . .
. X .|2 . .|. . .
X 2 X|X . .|X . .
-----+-----+-----
1 X X|. . .|X . X
. . .|. . .|X . 1
. . .|. A .|. . .
-----+-----+-----
. X 2|. . .|. . .
X X X|X X 1|. . .
. . .|. . .|. A .
User avatar
Pat
 
Posts: 3438
Joined: 18 July 2005

Postby RW » Sun Jun 04, 2006 1:57 pm

Pat wrote:- no, only 2 steps of "hidden singles"


Oops... I think I mentioned in another post that I spot hidden singles within boxes a lot easier than hidden singles in rows. Apparently this is a good example of that...

RW
RW
2010 Supporter
 
Posts: 1000
Joined: 16 March 2006

Postby gsf » Mon Jun 05, 2006 4:39 pm

ab wrote:Here it is:
1. Find all locked pairs (type 1), but pretend they do not exist.
2. Make eliminations based on the locked pairs found in step 1.
3. Find all hidden singles, but pretend they do not exist.
4. Place the hidden singles found in step 3 and perform eliminations.
5. Increment step counter and repeat from step 1.


tarek wrote:You are right gsf in saying that a box line elimination is possible.
however, I actually wouldn't go for a box-line elimination if there is
an obvious hidden single there.....I would place the hidden single only

[If two deductions can be made based on one cell then I would go for
the more obvious or easier deduction first, if that was the hidden
single then the step ends there. If it was the box-line elimination
then you are correct in your assumption].

I think that the setting should be at >1 per line per box because I
consider the hidden single EASIER than box-line elimination.


the point I hope is not to do what any particular user would do
but to do what the rules say
ab's ulterior rules don't have provisions for "more obvious"
Code: Select all
        (1) locked pairs
        (2) commit
        (3) hidden singles
        (4) commit
        (5) increment step count


tarek wrote:You are solving it Pat in the method which this thread does not employ
(we employ ab's method, Hidden & box-line at the same time).... (Pat's
method, is exhausting hidden singles before going to type one box-line
eliminations)......

The puzzle you mentioned solves in 12 steps(ab's method), 11 out of
these steps you can make box-line eliminations out of which 6 should
put more placements than if left for hidden singles alone......


even though its been spelled out there is still confusion
ab's rules are not exacty "Hidden & box-line at the same time"

I've was confused too until I came up with an expression syntax
(that can be input to my solver) that defines what a "step" is

(you can substitute any constraint names you like in the following
the operations on the constraints are what's important)

define a move to be one or more placements or eliminations
moves are identified by applying a constraint for a puzzle position
moves are committed by actually making the placements or eliminations
a concatenation of constraint names specifies a left to right order on constraint application

Code: Select all
C1...Ci...Cn


means:
Code: Select all
  repeat:
        apply the constraints in order C1 ... Cn
        for the first Ci that has at least one move
                identify all Ci moves
                commit all Ci moves in one batch
                increment the step-count
                goto repeat
        end for
        if no Ci has moves then either
                the puzzle is solvable in step-count steps or
                the puzzle is not solvable (constrained) by the given constraints
        end if


let {...} treat groups of constraints as one constraint
Code: Select all
{C1...Cn}

means
Code: Select all
        identify all C1 moves
        ...
        identify all Cn moves
        commit all C1...Cn moves in one batch

(all constraints in the group are applied, not all need produce moves)

finally, let : within a group denote "commit":
Code: Select all
{C1:C2C3}

means
Code: Select all
        identify all C1 moves
        commit all C1 moves
        identify all C2 moves
        identify all C3 moves
        commit all C2 and C3 moves


the inferior and ulterior threads involve these constraints
Code: Select all
        F       naked singles
        N       hidden singles
        B2      locked pairs (box-line)


so the ulterior step count expression (ab's rules) is
Code: Select all
        {B2:N}


and the inferior step count expression is
Code: Select all
        {FN}

the {FN} expression captures all of the posted inferior counts
{B2:N} captures the ulterior counts except the cases where F and/or N may also be identfied by B2
for reproducability it seems that B2 should identify singles if those singles match the
B2 constraints
gsf
2014 Supporter
 
Posts: 7306
Joined: 21 September 2005
Location: NJ USA

PreviousNext

Return to General