Draco wrote:Here's one that my solver rates as pretty hard, SE gives a 7.3 and both need several forcing chains (16 if I recall SE's list correctly, 20 for my solver if I let it go after short chains first, which I tend to do) to solve.

Any one able to bring some brevity to this situation, please chime in! ...

- Code: Select all
`.......91..48.1....6...534....4...89....5....29...6....872...3....6.75..62.......`

378 357 235 | 37 23467 234 | 2678 9 1

379 37 4 | 8 23679 1 | 267 2567 2567

1789 6 1289 | 79 279 5 | 3 4 278

----------------+-----------------+----------------

137 1357 1356 | 4 237 23 | 1267 8 9

13478 1347 1368 | 1379 5 2389 | 1267 1267 2367

2 9 138 | 137 378 6 | 147 157 3457

----------------+-----------------+----------------

5 8 7 | 2 149 49 | 1469 3 46

1349 134 139 | 6 1389 7 | 5 12 28

6 2 139 | 5 13489 3489 | 14789 17 478

As I mentioned in my reply to hobiwan earlier in this thread, I've been working on a chain search algorithm that would produce better forcing or contradictions chains/nets. What I've got is slow and, in limited testing, always at least as well or better than my old method. It can also be MUCH slower. In the case of the puzzle above, it reduced the # of chains needed from 16 or so to 2!

----- First Chain ------

r1c1=8 r3c9=8 r8c9=2 r8c5=8

r1c1=8 r9c7=8 r7c7=9

r9c7=8 [r9c6<>8] r5c6=8

------ Second Chain ------

r1c7=8 r1c5=6 r1c6=4 r7c6=9

------ Cancellations ------

r5c6<>9, r7c5<>9, r8c5<>9

------ Chain Ends ------

STSS then

------ First Chain ------

r3c3=1 r9c3=3 r6c3=8 r5c6=8

r3c3=1 r8c3=9

r3c3=1 r1c3=2 r1c6=4

r9c3=3 [r8c1<>3 r8c2<>3 r8c3<>3] r8c5=3

r6c3=8 [r6c5<>8] + r8c5=3 [r6c5<>3] = r6c5=7

r8c5=3 [r4c5<>3] + r6c5=7 [r4c5<>7] =

r4c5=2[r1c6<>2 + r5c6<>2] =

r4c6=2------ Cancellations ------

r3c3<>1

STSS will solve. More work to do on this new search code, but it is looking very promising!

Cheers...

- drac