The New Sudoku Players' Forum

by **ronk** » Wed Sep 27, 2006 5:49 pm

ravel wrote:I keep the original version that Tarek posted ...

At gsf's suggestion, a recent exercise showed that the Sudoku Explainer rating for each of 9 permutations of Ocean's SE=10.0 puzzle was unchanged.

So I'm curious if the RMS (ravel's minimal step

) rating of the permutation I posted is anywhere near the 20 steps of tarek's original.

by **ravel** » Thu Sep 28, 2006 7:09 am

ronk wrote:So I'm curious if the RMS (ravel's minimal step ) rating of the permutation I posted is anywhere near the 20 steps of tarek's original.

The reason, that the step count differs for isomorphic puzzles is, that if there are several possible eliminations, that all lead to the same number of candidate reductions by applying basic methods (if the elimination is made), in the original version i select the next one in the order it has in the puzzle.
When the step count is greater 10 or 15 resp. i repeat it at least 5 or 9 times, where now the elimination (under the "best possible" ones) is selected randomly.
For this puzzle i did it 28 times in the meantime and did not find more than a second 20 step solution.

by **tarek** » Thu Sep 28, 2006 8:24 am

ravel wrote:The best solution found needed 20 steps (and no other with 20)

I didn't expect that, I thought it had potential but not "20 steps". The high step count is suggested from Suexrat & gsf's sudoku (older version) but not the extent of 20 steps :!:

Probably manual solvers could shed some light on that aspect.........

I will post the leftovers of that search later today to see if it brings out anything of significance.

My next search would be still with the diagonal pattern however it would be with 2 empty boxes, following this pattern (modified from an earlier post by ronk)

Code: Select all: . . * | . . * | . . . . * . | . * . | . . . * . . | * . . | . . . -------+-------+------- . . * | . . * | . . * . * . | . * . | . * . * . . | * . . | * . . -------+-------+------- . . . | . . * | . . * . . . | . * . | . * . . . . | * . . | * . .

I suspect that the results would be disappointing because of the limited clue numbers (21 max).......

tarek

by **ravel** » Thu Sep 28, 2006 10:34 am

Eioru,

in your last list 3 is minimal to 4, 5 is minimal to 6 and 11 is minimal to 10.
Nr 3 needed 10 steps, nr 5 5 steps, and nrs 7 and 9 4 steps each.

I will update the list later, when i have more time.

by **Eioru** » Thu Sep 28, 2006 1:18 pm

1..2..3......4......5..6..72..1..8...9.....5...1..5..68..3..6......8......7..9..4
1..2..3......4......5..1..67..5..1...8.....9...4..2..82..9..7......8......6.....4
1..2..3......4......5..6..23..7..4...5.....8...6.....59..3..7......9......8..1..6
1..2..3......4......5..6..23..7..4...5.....8...6..8..59..3..7......9......8.....6
1..2..3......4......4..5..67..6..8...3.....9...6..7..52..9..1......8......5..4..8

by **tarek** » Fri Sep 29, 2006 4:35 am

Here are the leftovers from my last search......
As they were too many, I'll post puzzles according to this rule:

Code: Select all: Suexrat9>=500 OR (Suexrat9>=400 AND SE>=9.4)

Group1 (1st move advanced x-cycle technique)

Code: Select all: 700009800090030040005600001006000009040000030200000100100008700050040010007200005 900006800010030040007500003005000006040000030600000200200008100070040050009100000 300005800040030060007100002009000008030000040100000900600009000050040080002800007 100006800040030020006900003005000004020000030600000900900008100070040000001700005 900006100080030040005700003007000006060000080300000200200001900050040010009300005 800001900020030040000600000007000002040000030500000600600009800070040050008100009

Group2 (1st move Hidden double)

Code: Select all: 000007600040030080002500003007000002030000070900000100600008900070040060009100005 100004800070030020006900003005000007020000030600000900900008100000040060001200005 900004300040030020005900001007000004020000010800000600600003100050040070009800005 000009400080030050001700006007000002020000080300000100400001600050040030009200000 800001700020030050000600009003000004060000020100000800600008000050040030009700005

Each group is sorted according to SE then Suexrat9

tarek

by **ravel** » Fri Sep 29, 2006 1:44 pm

Eioru,

puzzles 1 and 3 of the last list needed 4 steps.

Tarek,

again a heavy weight list, though not so outstanding as the 20-stepper. Since my program has not finished yet, i only can tell the steps after the first run at the moment. So for those with a questionmark lower ratings are to expect at the end.
10(?),5,7,9,5
12(?),9,11(?),5,9

I will update the list next week.

by **ravel** » Mon Oct 02, 2006 8:51 am

List updated.
the final ratings for Tarek's list are
10,5,7,9,5,3
9,9,10,5,9

I also added a top 10 list (for the puzzles in the "hardest list") according to gsf's rating.

by **Eioru** » Tue Oct 03, 2006 9:46 am

Are these two the hardest puzzles of 17-clue and 18-clue?

Code: Select all: ..13...8. 6...9.... ......... .23....1. ....7.4.. .8.5..... 4.....9.7 ...2.5... ......6..

Code: Select all: ...1..... .......27 6..54.... 89...2... .7....5.. ......4.. ......... .....8.9. 1.5.6....

by **ravel** » Tue Oct 03, 2006 11:05 am

Eioru wrote:Are these two the hardest puzzles of 17-clue and 18-clue?

Maybe the 17-clue, ER 9.5, gsfr 99570. But there should be many harder 18-clues (9.0/99419). Both had less than 4 BF steps.

by **gsf** » Tue Oct 03, 2006 5:05 pm

ravel wrote:
Eioru wrote:Are these two the hardest puzzles of 17-clue and 18-clue?
Maybe the 17-clue, ER 9.5, gsfr 99570. But there should be many harder 18-clues (9.0/99419). Both had less than 4 BF steps.

I happened to keep the 191634 of gordon's 22000000 18's that foil basic constraints
here are the top 10 from that list with # rating puzzle-ordinal commentary

Code: Select all: # 99691 #18968907 7.8...3.....2.1...5.........4.....263...8.......1...9..9.6....4....7.5........... # 99648 #18969101 7.8...3.....6.1...5.........4.....263...8.......1...9..9.2....4....7.5........... # 99547 #14231860 6..1...8..53.............4....8...6..9....7....24.........7.3.9....2.5..1........ # 99539 #19951018 .8....2.....9...4...1........51...9..6...8...4........5.....8.3....2.6..9..7..... # 99527 #12805432 .5.6..4..2...3..8..7........4.7..6......2..3..........3...8..1....5..7..1........ # 99517 #5556604 2.91........4..36........2...13..7...3..6........8.......5....182.......6........ # 99494 #432710 .1.3..4......62.8..........6....8.2.7..1..5............5.4..1..2.......6....3.... # 99488 #8328081 .41.......2.....5....7....3....21...5......7.......8....6.4.1.......52..8..3..... # 99477 #20256047 8..4.1.........43.......6..5.......1...37.......6...4...7.2.....3.5......46...... # 99476 #6422801 .3....27.5...1................3..46..7.2.....8.......11....8..5...4..3...2.......

by **Carcul** » Tue Oct 03, 2006 6:50 pm

Eioru wrote:Are these two the hardest puzzles of 17-clue and 18-clue?

For the 18 clues puzzle:

Code: Select all: *-----------------------------------------------------------------------* | 379 38 3789 | 1 2 3679 | 3689 5 4 | | 3459 1345 1349 | 3689 389 369 | 1369 2 7 | | 6 1238 123789 | 5 4 379 | 1389 13 138 | |------------------------+-----------------------+----------------------| | 8 9 1346 | 3467 5 2 | 137 1367 136 | | 234 7 12346 | 34689 389 13469 | 5 1368 29 | | 235 12356 1236 | 36789 3789 1369 | 4 13678 29 | |------------------------+-----------------------+----------------------| | 23479 2348 234789 | 2347 137 5 | 12378 13467 1368 | | 2347 2346 23467 | 2347 137 8 | 1237 9 5 | | 1 2348 5 | 23479 6 349 | 2378 347 38 | *-----------------------------------------------------------------------*

[r5c6]=4=[r9c6](-4-[r9c2])-4-[r9c8]=4=[r7c8]-4-[r179c2]-2,3,8-[r3c2]
(-1-[r39c9]-3-[r4c9])-1-[r3c8](-3-[r456c8]=3=[r4c7])-3-[r9c8]-7-
-[r456c8])=7=[r4c7],

and so r9c6<>4 which solves the puzzle.

Carcul

by **ravel** » Wed Oct 04, 2006 9:39 am

Carcul, thanks for the new solution riddle.

The ten puzzle collection filtered by gsf brought the first 18-clues to my list (7,4,4 steps), thanks.
But it also showed again, how different the ratings are (RMS is Ron's acronym for ravels minimal steps - the number of brute force steps my program needed). I did not calculate the minimum number for those with less 4 steps.

Code: Select all: nr gsfr ER RMS ----------------------------- #18968907 99691 9.5 #18969101 99648 9.5 #14231860 99547 9.2 #19951018 99539 9.1 4 #12805432 99527 9.2 4 #5556604 99517 7.8 #432710 99494 7.1 #8328081 99488 9.0 7 #20256047 99477 7.2 #6422801 99476 7.3

For the 7th puzzle i saw a solution with a 5-cell forcing chain (r6c4=2) and advanced coloring (in 7).

by **Carcul** » Wed Oct 04, 2006 11:43 am

Ravel wrote:Carcul, thanks for the new solution riddle.

I don't see any riddle.

Carcul

by **ronk** » Wed Oct 04, 2006 1:02 pm

Carcul wrote:
Ravel wrote:Carcul, thanks for the new solution riddle.

I don't see any riddle.

Lots of laughs, but I have to agree with ravel. When an expression has three multiple inferences ... and the leftmost and rightmost cells don't even share a unit ... it is a riddle to most of us.

The New Sudoku Players' Forum

The hardest sudokus