The New Sudoku Players' Forum

by **ronk** » Sun Jul 15, 2012 4:58 pm

daj95376 wrote:Hmmm. My analyzer didn't find the <3457> pattern. It may have been blocked by finding the <1268> pattern. Something to check.

Perhaps because it's a 5 digit fish <34579> instead of 4. However, there are 19 base units instead of 20.

by **daj95376** » Sun Jul 15, 2012 5:27 pm

ronk wrote:Perhaps because it's a 5 digit fish <34579> instead of 4. However, there are 19 base units instead of 20.

Aaaaagh!!! I missed the (clearly shown) additional Truth set. Thanks for saving me a lot of review work. FWIW: I only check for 2/3/4-template patterns with my analyzer. My solver has the capability to resolve 2/3/4/5/6/7-template patterns; but, that's overkill since (I think) eleven said that 5-template patterns are the max.

Have you found Multi-Fish in any of the remaining 26 puzzles in file "03 NN ..."? If so, then no need for me to continue searching.

Regards, Danny

by **ronk** » Sun Jul 15, 2012 6:51 pm

daj95376 wrote:Have you found Multi-Fish in any of the remaining 26 puzzles in file "03 NN ..."? If so, then no need for me to continue searching.

Including the one we've just discussed, I've noted these:

Code: Select all: 1....6.......8.2....97....5.3.9....4..5....9.....2.1....4.....7.9.3...4.8.....6.. ;1582;elev;567;0-rank 19T 5-dig 4-row-fish .......8....1.9..66...2...4.7...8.9.5..........4.3...73.2.....5..5.6.3...1.....7. ;3218;elev;997;0-rank 18T 4-dig 5-col fish with box cover ....5...94..1..2.......7...2.8.1.6....62......7......3.3..9..5.8..........28..4.. ;4971;elev;1943;0-rank 15T 5-dig 4-row fish ........9..71...6.....3.1.42..6...1...5.......8...43..5......2.6.17......9...8... ;1690;elev;301;0-rank 15T 4-dig 4-col fish .2...67..4...8......9.....1..1.....8...2..6...6...3.5......5.2..7.6..3..9...4.... ;4582;elev;2279;0-rank 14T 4-dig 5-row fish 1...5......7..9..6.983.......67....8.8.......5...4..2.....3.1...3.9....7.....2.4. ;3173;elev;1405;0-rank 15T 4-dig 4-col fish ...4..7..4...8...6..9..2.......3...83..8..4...18....5..9.....1.7..6....3..2..5... ;2926;elev;L294;0-rank 18T 5-dig 4-row fish 1..4..7....7.8...6.9.......2..3......4...71....5.4..2..3.9..4......6...8..2....5. ;7417;elev;H193; (listed fish was bogus)

Think I looked at all the puzzles in this file, but might have easily missed something.

[edits: 1) add puzzle #1690; 2) ; #1690 was 19T 5-dig 4-row; commented #7417, listed fish was actually #1690 again]

by **pjb** » Mon Jul 16, 2012 1:22 am

When I examined out of interest how my newly created multi-fish implementation (here) handled the '03 NN nothing special hard' set of puzzles, I was surprised how my results differed from those posted above by Ronk. I found 8 with a muli-fish pattern, but they were quite different. Not counting #1690, I have 3 not in Ronk's list and failed to find 3 of his. Even the ones in both lists have different solutions. I only look for row truths, since it seems to me that in these simple row-column multi-fishes there is always a complementary column solution and vice versa. My results are:

Code: Select all: 1....6.......8.2....97....5.3.9....4..5....9.....2.1....4.....7.9.3...4.8.....6.. ;1582;elev;567;0-rank 20T 4-dig 5-row-fish .......8....1.9..66...2...4.7...8.9.5..........4.3...73.2.....5..5.6.3...1.....7. ;3218;elev;997; - ....5...94..1..2.......7...2.8.1.6....62......7......3.3..9..5.8..........28..4.. ;4971;elev;1943; - ........9..71...6.....3.1.42..6...1...5.......8...43..5......2.6.17......9...8... ;1690;elev;301;0-rank 19T 4-dig 5-row fish 1..4..7....7.8...6.9.......2..3......4...71....5.4..2..3.9..4......6...8..2....5. ;7417;elev;H193; - .2...67..4...8......9.....1..1.....8...2..6...6...3.5......5.2..7.6..3..9...4.... ;4582;elev;2279;0-rank 20T 4-dig 5-row fish 1...5......7..9..6.983.......67....8.8.......5...4..2.....3.1...3.9....7.....2.4. ;3173;elev;1405;0-rank 19T 4-dig 5-row fish ...4..7..4...8...6..9..2.......3...83..8..4...18....5..9.....1.7..6....3..2..5... ;2926;elev;L294;0-rank 20T 4-dig 5-row fish 98.7.....7...8.6....5..4...6..3..9...9.....2...4..1..6.3.8..7.......2.1.........5;227;GP;H29;0-rank 15T 4-dig 4-row fish .....6.8.4..7....3.9..3.....1..4...73..1.......8..52...3.9....4......5....2....6.;4891;elev;1905;0-rank 20T 4-dig 5-row fish .......89.5.1.....6....31...7...16......9..2...8.....4..4.2....7..5..3...6...7...;3329;elev;991;0-rank 20T 4-dig 5-row fish ........9..71...6.....3.1.42..6...1...5.......8...43..5......2.6.17......9...8...;1690;elev;301;0-rank 19T 4-dig 5-row fish

This just goes to show that there is more than one way to skin a cat!
pjb

by **champagne** » Mon Jul 16, 2012 8:20 am

pjb wrote:When I examined out of interest how my newly created multi-fish implementation (here) handled the '03 NN nothing special hard' set of puzzles, I was surprised how my results differed from those posted above by Ronk. I found 8 with a muli-fish pattern, but they were quite different. Not counting #1690, I have 3 not in Ronk's list and failed to find 3 of his. Even the ones in both lists have different solutions. I only look for row truths, since it seems to me that in these simple row-column multi-fishes there is always a complementary column solution and vice versa. My results are:

This just goes to show that there is more than one way to skin a cat!
pjb

I did the same using my last code.
This is a code just activated for the 4 digit pattern and not yet in use (I did not make the regression test so far)

I found 10 puzzles with a multi_fish pattern and missed the 2 above in ronk's list

I'll study later the results

Hidden Text: Show

champagne

by **champagne** » Mon Jul 16, 2012 1:59 pm

Here the list of puzzles where multi_fish has been found.

I had not yet the time to check in detail the results, but from that file, all have a row and a columns based SLG

Hidden Text: Show

[ronk-moderator edit: To retain some mystery for players, base digits deleted.]

by **champagne** » Mon Jul 16, 2012 6:57 pm

EDIT

the list of puzzles with truths/links found by my solver;

Compared to ronk's list
puzzle 3218 is missing likely due to a pattern not seen by my solver
puzzle 7417 is missing for an unknown reason

puzzle 3173 from pjb's list is missing likely for the same reason as for puzzle 7417

I am not sure to have the time to go deeper the next three days

Hidden Text: Show

by **ronk** » Tue Jul 17, 2012 7:34 pm

champagne, I'm happy to see your truth\link data in the xsudo format,

... but what happened to the plan to "retain some mystery for players?"

by **champagne** » Wed Jul 18, 2012 6:08 am

ronk wrote: I'm happy to see your truth\link data in the xsudo format,

as you can see, even discouraging comments ( dyslexia !!) can be taken positively.
An I am glad to learn that now it's ok for direct pasting in XSUDO

ronk wrote:
... but what happened to the plan to "retain some mystery for players?"

I have seen that comment. My understanding has been that something in my post was a source of trouble for a quick reading.

May be the reason is that I am very far from your view regarding the current the situatio(n, although I gave the solution of my solver in hidden mode.)

For me we are at the following point.

Three of us tested the same file, at least 2 with a program.

We see significant deviations in the results, which means for the programs either a lack of design or bugs in the code.

I consider that exchanging the results is a huge plus to improve the coding.

I did not yet dig in the results and will not do it before next Friday, but I am sure I would appreciate to have at that time the results where my solver fresh code failed. This saves a lot of time.

Players have the possibility to keep the hidden part of the post hidden,
but our work is also the source of sample files in which they can look.

At the end, we should find a way to be positive for everybody.

by **David P Bird** » Wed Jul 18, 2012 10:34 am

Solving manually, the difficulty with these multi-fish is identifying the digit set to use. Sometimes they stand out but at others they can be tough. With I'm practice I'm getting better at this, but I'm far from having developed a foolproof algorithm.

Remembering that I'm just concerned with how may times a cell is covered (by row and column) and using a colouring scheme to show this, here's more or less what I do:

Identify a "makeweight" digit as being one that occurs most frequently as a given, and often has been solved somewhere in the preliminary eliminations.

The objective is to partition the 8 remaining digits into sets where their givens frequently occur as peers in the same rows and columns. In some case this can be quite straightforward, but when there are several compromise options available it's useful to break down the search into sub-stages.

First look for rectangles of givens composed of two digits as these are almost certain to be in the same partition. If there are two such rectangles, try using the four digits together as one partition.

Failing that, start with 3 digits which have a good peer count and cover the rows and columns containing their givens. Now consider which of the remaining digits can be most fruitfully added to that set so that 9 rows and columns in total will cover as many of their givens as possible.

Assuming a good compromise has been reached check to see if a balance has been reached. Here the multi-fish methods used diverge:
For the truth and link set approach I believe this would involve forming truth sets for one row/column direction and counting the link sets needed to complete the coverage.
For the Shark approach, when 9 rows and columns have been covered, the minimum number of truths that must be in the doubly covered cells must equal the maximum number of truths that can be held in the completely uncovered cells. (When the row/column count, T, doesn't equal 9. the formula to balance is Min = Max - N*(9-T) where N is the number of digits in the set.)

If a balance has been achieved then a multi-fish has been found, if not, add the makeweight digit to the digit set and re-check. If this balances the complementary 5 digit Multifish has been found.

Using this method I've manually found multi-fish for all the puzzles listed except for 3218 which is still defeating me. Other than that, the toughest one was 3173.

I'm still working through those where no multi-fish has been found so far. Four of the five I've tried up to now I would class as Almost patterns as their balances are out by just one.

by **ronk** » Wed Jul 18, 2012 11:23 am

champagne wrote:Three of us tested the same file, at least 2 with a program.

We see significant deviations in the results, which means for the programs either a lack of design or bugs in the code.

I consider that exchanging the results is a huge plus to improve the coding.

I have no program but use Xsudo which, as stated before I think, sometimes helps and sometimes doesn't. For the 12 puzzles under discussion, Xsudo and I found a 0-rank multi-fish for all without knowing the truth\link sets or even the base digits. Before pjb's and your posts, I only had 8 of the 12.

According to the most recent posts, you have fish for #1263 and #4971 where pjb does not. I will post the fish for #3218 and #7417, which neither of you seem to have, at some unknown future date.

champagne wrote:I am glad to learn that now it's ok for direct pasting in XSUDO

As the software writer, didn't you test this yourself?

champagne wrote:as you can see, even discouraging comments ( dyslexia !!) can be taken positively.

Since I wrote "Dyslexia?

", with a question mark and a smiley face, your "dyslexia!!" is deceptive IMO but, since you brought it up, would you please write a response that would have been acceptable to you?

by **ronk** » Wed Jul 18, 2012 12:28 pm

David P Bird wrote:Using this method I've manually found multi-fish for all the puzzles listed except for 3218 which is still defeating me. Other than that, the toughest one was 3173.

David, good show! Please post your multi-fish for #7417. champagne and pjb apparently have not found this one yet.

by **daj95376** » Wed Jul 18, 2012 3:57 pm

I would like to thank champagne for his recent post of solutions. It saved me from beating my head against the wall trying to take my Template Analyzer results and derive Multi-Fish from it. I was able to find solutions for many of the puzzles listed by others, but there were several that weren't successful. Thanks to one of champagne's solutions, I was able to determine that my approach used an over-constrained premise that prevented the Multi-Fish in these additional puzzles from being found. Oh Well, it was fun trying.

Regards, Danny

Addendum:

For the record, my screwed up approach found Multi-Fish for these puzzles:

Hidden Text: Show

by **David P Bird** » Wed Jul 18, 2012 7:50 pm

ronk wrote:David, good show! Please post your multi-fish for #7417. champagne and pjb apparently have not found this one yet.

Not have I! Somehow I didn't add a "solved" tag for it in my list. When I saw your post I thought I'd just polish it off, but it didn't succumb to my methods.

Here's a list of those that I've found, plus the first four that that I class as Almost Sharks. Be warned, this list shows the digit sets and the cover houses used for each puzzle.

Hidden Text: Show

[Edit] The hidden text has been revised to show my results for the full set of puzzles. 13 of the 16 puzzles with no Multifish have Almost Sharks.

by **champagne** » Thu Jul 19, 2012 4:42 am

waking up early this morning, I have small time to analyse the situation.

I restarted from the list published in the "hardest new thread" (same content as 03 NN but sorted differently)

in that list, puzzle 7417 is said to have nearly no chance to give a multi-fish pattern.

In fact, my solver finds no floor group of size 4 or 5 leading to eliminations (danny should have the same results in template search)
So my solver did not try to build a multi-fish in that case

It will be very very interesting to see ronk's findings for that puzzle.

Another interesting point is that, as expected, nearly all the bottom part of the list can now be erased.

The best chances to find new multi fish in that group are
puzzle 1690
puzzles 12120
puzzle 4993

and may be also puzzles 1102 14596 3909 and 35

after the next update of the data base of potential hardest, I'll work on a new list but puzzles 4024 3411 889 7417(????) 1362 have high chances to stay there.

The New Sudoku Players' Forum

Exotic patterns a resume

Re: Exotic patterns a resume

Re: Exotic patterns a resume

Re: Exotic patterns a resume

Re: Exotic patterns a resume

Re: Exotic patterns a resume

Re: Exotic patterns a resume

Re: Exotic patterns a resume

Re: Exotic patterns a resume

Re: Exotic patterns a resume

Re: Exotic patterns a resume

Re: Exotic patterns a resume

Re: Exotic patterns a resume

Re: Exotic patterns a resume

Re: Exotic patterns a resume

Re: Exotic patterns a resume