The New Sudoku Players' Forum

Website · by **denis_berthier** » Fri Jan 15, 2021 5:31 am

Mauriès Robert wrote:
denis_berthier wrote:But the discussion was not between whips and antitracks; it was between a whip and a pair of conjugated tracks. You keep repeating that the latter have an easier to find starting point. I've shown that this claim is false.

I didn't say anything like that, read what I wrote again.

You have written in every possible thread that the starting point for a whip was its target, but it was easier to start with conjugated track. I'm certainly not going to search all this site for that.

But I'm glad to hear that, at last, you admit it's a false claim.

by **Mauriès Robert** » Fri Jan 15, 2021 2:57 pm

denis_berthier wrote:You have written in every possible thread that the starting point for a whip was its target, but it was easier to start with conjugated track. I'm certainly not going to search all this site for that.
But I'm glad to hear that, at last, you admit it's a false claim.

Indeed, don't search all over the site you wouldn't find such a statement from me. What I have been able to say about a target concerns the anti-tracks I compare to whips and braids and not the conjuguated tracks.
I said that with an anti-track P'(E) the target is found as a consequence of its development among all the candidates who see E, whereas for whips and braids it is not, the target is chosen as a starting point. On this last point, I have admitted being wrong, the starting point for the whips/braids and the antitracks are the same, we have agreed on this, which for me implies that the target is found with the whips/braids as a consequence of their development among the candidates who see the candidate on the left of the variable CSP V1. But as you pointed out to me before, we are off-topic, the thread being devoted to forcing-T&E to which I have nothing to add.
Having said that, I have some questions to ask you, which I will do in private because it is not a question of debating but of understanding.

Cordialy
Robert

Website · by **denis_berthier** » Fri Jan 15, 2021 3:18 pm

Mauriès Robert wrote:I have some questions to ask you, which I will do in private because it is not a question of debating but of understanding.

Questions of understanding are probably more interesting to the silent readers of this forum than the usual pointless debates.

by **Mauriès Robert** » Fri Jan 15, 2021 4:52 pm

denis_berthier wrote:
Mauriès Robert wrote:I have some questions to ask you, which I will do in private because it is not a question of debating but of understanding.

Questions of understanding are probably more interesting to the silent readers of this forum than the usual pointless debates.

Ok, so I'm going to create a topic to ask you questions with the objective of better understanding your theories.
Robert

Website · by **denis_berthier** » Tue Feb 16, 2021 11:49 am

.
Forcing{3}-T&E

Forcing-T&E is very inelegant and it is not a panacea.
In the following example (Ref.: http://forum.enjoysudoku.com/sudokus-with-a-nice-form-t38506-15.html#p300739)

JPF wrote:Here is an other example with the same pattern:
Code: Select all
. . 1 | . 2 . | 3 . . . 4 . | . 1 . | . 5 . 6 . . | 3 . 5 | . . 1 -------+-------+------- . . 6 | . . . | 7 . . 2 1 . | . . . | . 6 3 . . 5 | . . . | 1 . . -------+-------+------- 5 . . | 1 . 8 | . . 9 . 9 . | . 6 . | . 3 . . . 8 | . 9 . | 5 . . ED=9.8/1.2/1.2

there is a solution in gW13, obtainable with the usual SudoRules rules and simplest-first strategy:

solution in gW13: Show

However, there is no solution using only Forcing-T&E as defined until now.
Starting from the same resolution state RS1 as in the gW13 solution, one can get the following eliminations:

applying Forcing-T&E: Show

FORCING-T&E(S) applied to bivalue candidates n3r6c6 and n3r9c6 :
===> 0 values decided in both cases:
===> 3 candidates eliminated in both cases: n3r9c2 n4r9c6 n7r9c6

Code: Select all: CURRENT RESOLUTION STATE: 789 5 1 46789 2 4679 3 4789 4678 3789 4 2379 6789 1 679 2689 5 2678 6 278 279 3 478 5 2489 24789 1 3489 38 6 2489 348 1 7 2489 5 2 1 479 4789 5 479 489 6 3 34789 378 5 246789 3478 234679 1 2489 248 5 2367 2347 1 347 8 246 247 9 1 9 247 5 6 247 248 3 2478 347 267 8 247 9 23 5 1 2467

FORCING-T&E(S) applied to bivalue candidates n8r5c4 and n8r5c7 :
===> 0 values decided in both cases:
===> 1 candidates eliminated in both cases: n8r3c7

Code: Select all: CURRENT RESOLUTION STATE: 789 5 1 46789 2 4679 3 4789 4678 3789 4 2379 6789 1 679 2689 5 2678 6 278 279 3 478 5 249 24789 1 3489 38 6 2489 348 1 7 2489 5 2 1 479 4789 5 479 489 6 3 34789 378 5 246789 3478 234679 1 2489 248 5 2367 2347 1 347 8 246 247 9 1 9 247 5 6 247 248 3 2478 347 267 8 247 9 23 5 1 2467

And that's all. No more Forcing-T&E can be applied successfully.

I therefore coded a generalisation of Forcing-T&E, Forcing{3}-T&E where trivalue cells can be used as starting points instead of bivalue ones. Needless to say, this is still more inelegant than starting with bivalue cells, but I wanted to try.
The rules are obvious: start a separate T&E procedure for each of the 3 candidates in a trivalue cell; any candidate deleted (resp. asserted as a value) in the three procedures can be deleted (resp. asserted) in the solution.
It works in the present case, providing a solution in six steps (still starting from RS1):

Solution with Forcing{3}-T&E: Show

I haven't yet tried to mix the above two variants of Forcing-T&E.

by **creint** » Tue Feb 16, 2021 8:56 pm

denis_berthier wrote:.
there is a solution in gW13, obtainable with the usual SudoRules rules and simplest-first strategy:
[hidden=solution in gW13]***********************************************************************************************
*** SudoRules 20.1.s based on CSP-Rules 2.1.s, config = gW+SFin
*** Using CLIPS 6.32-r779
*** Download from: https://github.com/denis-berthier/CSP-Rules-V2.1
***********************************************************************************************
singles ==> r9c8 = 1, r8c1 = 1, r8c4 = 5, r5c5 = 5, r4c9 = 5, r4c6 = 1, r1c2 = 5
177 candidates, 1002 csp-links and 1002 links. Density = 6.43%
Code: Select all
RESOLUTION STATE RS1: 789 5 1 46789 2 4679 3 4789 4678 3789 4 2379 6789 1 679 2689 5 2678 6 278 279 3 478 5 2489 24789 1 3489 38 6 2489 348 1 7 2489 5 2 1 479 4789 5 479 489 6 3 34789 378 5 246789 3478 234679 1 2489 248 5 2367 2347 1 347 8 246 247 9 1 9 247 5 6 247 248 3 2478 347 2367 8 247 9 2347 5 1 2467

solution in gW13:
How fast should this solve? I takes a while on my pc > 3 min.

Code: Select all: whip[2]: r5n8{c7 c4} - b2n8{r1c4 .} ==> r3c7 ≠ 8

It looks its missing r2c4 from the false set, shouldn't it be:

Code: Select all: whip[2]: r5n8{c7 c4} - b2n8{r12c4 .} ==> r3c7 ≠ 8

Code: Select all: r5n8{c4 c7} - r8n8{c7 c9} - r2n8{c9 c1} - c2n8{r3 r6} - r4c2{n8 n3} - c1n3{r6 r9} - c6n3{r9 r6} - r6n6{c6 c4} - b5n2{r6c4 .} ==> r4c4 ≠ 8

It looks like c2n8{r3 r6} is skipping 8r4c2, at that point you could see it as and OR chain: 8r4c2 -> -5r4c4 OR 8r6c2 and then the remaining logic. Does it skip that because it can directly see the target? Why does the chain already know which target it is excluding?
Also there is no mention that -8r2c36 is used from previous steps at step: r2n8{c9 c1}
Why is there . inside b5n2{r6c4 .}, shouldn't is be b5n2{r6c4 r4c4}? Following the logic -2r6c4 what should follow: any placement that could be made that exclude 8 in r4c4?

Code: Select all: g-whip[11]: b4n3{r6c2 r456c1} - r2n3{c1 c3} - r7n3{c3 c5} - c6n3{r9 r6} - r6n6{c6 c4} - b5n2{r6c4 r4c4} - b5n9{r4c4 r5c456} - c3n9{r5 r3} - c7n9{r3 r2} - c7n6{r2 r7} - c2n6{r7 .} ==> r9c2 ≠ 3

b4n3{r6c2 r456c1} should be b4n3{r6c2 r46c1} because r5c1 is already filled with 2.

Website · by **denis_berthier** » Wed Feb 17, 2021 2:29 am

creint wrote:solution in gW13:
How fast should this solve? I takes a while on my pc > 3 min.

On my Mac, it takes 3.5 mins. Notice that only an infinitesimal proportion of puzzles need so long chains.

As for the rest of your questions, before commenting what whips, g-whips... are or how they should look, read their definitions.
z and t- candidates are not written (not "skipped"); they are not part of the pattern.

by **Leren** » Wed Feb 17, 2021 9:42 am

Couldn't resist trying Hodoku for it's time to solve this one. Essentially no time, but uses 6 long forcing chains and one Brute force (T&E ) for a large "score" of 18,194, but a long way from 3 minutes. Leren

Website · by **denis_berthier** » Wed Feb 17, 2021 9:53 am

Leren wrote:Couldn't resist trying Hodoku for it's time to solve this one. Essentially no time, but uses 6 long forcing chains and one Brute force (T&E ) for a large "score" of 18,194, but a long way from 3 minutes. Leren

Using T&E, any 9x9 puzzle can be solved in a tiny fraction of a second. The more resolution rules you use, the longer it will take to have a pattern-based solution.
Hodoku using brute force (i.e. Guessing, worse than T&E) means it is unable to solve it using only resolution rules.

by **DEFISE** » Mon May 17, 2021 10:13 am

Hi Denis,
by reading this thread "Forcing T&E" from the beginning, I found a beautiful “pearl” !
But, be reassured, it does’nt come from you but from Robert, in his January 13 post.
He found an anti-track to eliminate 3r4c6. But it turns out that this candidate is part of the solution.

Website · by **denis_berthier** » Mon May 17, 2021 12:52 pm

DEFISE wrote:Hi Denis,
by reading this thread "Forcing T&E" from the beginning, I found a beautiful “pearl” !
But, be reassured, it does’nt come from you but from Robert, in his January 13 post.
He found an anti-track to eliminate 3r4c6. But it turns out that this candidate is part of the solution.

Hi François
Manual solutions are more prone to such errors.

by **Mauriès Robert** » Sat Mar 05, 2022 5:49 pm

DEFISE wrote:Hi Denis,
by reading this thread "Forcing T&E" from the beginning, I found a beautiful “pearl” !
But, be reassured, it does’nt come from you but from Robert, in his January 13 post.
He found an anti-track to eliminate 3r4c6. But it turns out that this candidate is part of the solution.

Thank you François for pointing out this "silly" mistake on my part. I had not seen your intervention until today, hence my late reply!
I have corrected my text with another example that illustrates my text without changing the meaning.
Robert

Website · by **denis_berthier** » Tue Mar 08, 2022 6:56 am

.
The new example is as irrelevant to the topic under discussion as the deleted one. As I wrote:
But the discussion was not between whips and antitracks; it was between a whip and a pair of conjugated tracks. You keep repeating that the latter have an easier to find starting point. I've shown that this claim is false.

by **DEFISE** » Tue Mar 08, 2022 12:13 pm

Mauriès Robert wrote:
DEFISE wrote:Hi Denis,
by reading this thread "Forcing T&E" from the beginning, I found a beautiful “pearl” !
But, be reassured, it does’nt come from you but from Robert, in his January 13 post.
He found an anti-track to eliminate 3r4c6. But it turns out that this candidate is part of the solution.

Thank you François for pointing out this "silly" mistake on my part. I had not seen your intervention until today, hence my late reply!
I have corrected my text with another example that illustrates my text without changing the meaning.
Robert

Hi Robert,
Sorry for my persistence in looking for errors.
I forgot the context of your example, but I don't intend to read again this thread from the beginning.
To tell the truth, I am more interested in T&E than in forcing-T&E.

The New Sudoku Players' Forum

Forcing T&E

Re: Forcing T&E

Re: Forcing T&E

Re: Forcing T&E

Re: Forcing T&E

Forcing{3}-T&E

Re: Forcing{3}-T&E

Re: Forcing{3}-T&E

Re: Forcing T&E

Re: Forcing T&E

Re: Forcing T&E

Re: Forcing T&E

Re: Forcing T&E

Re: Forcing T&E

Re: Forcing T&E