The New Sudoku Players' Forum

[urhegyi]

Is it possible to break open this puzzle(eventually with T&E) in a human way?(random generated, first time I found such a high rating).

Code: Select all

.--------------------.--------------------.----------------------.
| 4     1258  1259   | 1239   138    289  | 12378  6      123578 |
| 156   1258  3      | 7      1468   2468 | 9      258    1258   |
| 169   7     1269   | 12369  1368   5    | 1238   238    4      |
:--------------------+--------------------+----------------------:
| 137   6     147    | 23     9      247  | 5      2378   12378  |
| 2     1345  14579  | 8      34567  467  | 1367   379    1379   |
| 3579  35    8      | 2356   3567   1    | 2367   4      2379   |
:--------------------+--------------------+----------------------:
| 1357  1345  1457   | 159    2      789  | 3478   35789  6      |
| 8     235   2567   | 4      567    679  | 237    1      23579  |
| 1567  9     124567 | 156    15678  3    | 2478   2578   2578   |
'--------------------'--------------------'----------------------'

4......6...37..9...7...5..4.6..9.5..2..8.......8..1.4.....2...68..4...1..9...3...



[denis_berthier]

(random generated, first time I found such a high rating).

That's quite surprising. SER 10.5 is very far from what one gets with the usual kinds of random generators (the usual upper bound is around 9.1).
Of course, it all depends on what you call random.

Is it possible to break open this puzzle(eventually with T&E) in a human way?

The puzzle is neither in T&E(1) nor gT&E(1) nor in T&E(Subsets, 1).
It is in T&E(W2, 1).
I doubt a resolution path in T&E(W2, 1) would be considered as a human way of solving.

Using regular chains, here's what I get:

Code: Select all

whip[9]: r6c2{n3 n5} - r5n5{c3 c5} - r5n3{c5 c2} - r8n3{c2 c9} - r8n5{c9 c3} - r1n5{c3 c9} - r1n7{c9 c7} - r8c7{n7 n2} - r8c2{n2 .} ==> r6c7≠3
g-whip[10]: c4n5{r9 r6} - r6c2{n5 n3} - c1n3{r6 r7} - c1n5{r7 r2} - b1n6{r2c1 r3c123} - c4n6{r3 r9} - r8c5{n6 n7} - r8c6{n7 n9} - r7n9{c6 c8} - c8n5{r7 .} ==> r9c5≠5
whip[13]: b9n9{r7c8 r8c9} - r8n3{c9 c2} - r6c2{n3 n5} - r5n5{c3 c5} - r8n5{c5 c3} - r8n2{c3 c7} - r9n2{c9 c3} - b7n6{r9c3 r9c1} - c3n6{r8 r3} - c4n6{r3 r6} - r6c7{n6 n7} - r1n7{c7 c9} - r1n5{c9 .} ==> r7c8≠3
   +----------------------+----------------------+----------------------+
   ! 4      1258   1259   ! 1239   138    289    ! 12378  6      123578 !
   ! 156    1258   3      ! 7      1468   2468   ! 9      258    1258   !
   ! 169    7      1269   ! 12369  1368   5      ! 1238   238    4      !
   +----------------------+----------------------+----------------------+
   ! 137    6      147    ! 23     9      247    ! 5      2378   12378  !
   ! 2      1345   14579  ! 8      34567  467    ! 1367   379    1379   !
   ! 3579   35     8      ! 2356   3567   1      ! 267    4      2379   !
   +----------------------+----------------------+----------------------+
   ! 1357   1345   1457   ! 159    2      789    ! 3478   5789   6      !
   ! 8      235    2567   ! 4      567    679    ! 237    1      23579  !
   ! 1567   9      124567 ! 156    1678   3      ! 2478   2578   2578   !
   +----------------------+----------------------+----------------------+

Note that I didn't try to use uniqueness.

[1to9only]

(random generated, first time I found such a high rating).

That's quite surprising. SER 10.5 is very far from what one gets with the usual kinds of random generators (the usual upper bound is around 9.1).
Of course, it all depends on what you call random.

"(the usual upper bound is around 9.1)" - this is true of the original SE created about 15 yrs ago when cpus were slow (compared to today) and nested chains were not checked during puzzle generation.

Must have been using one of my recent SEs... where I removed the nested chains restriction and a very high rating will occasionally be generated.

[denis_berthier]

.
Indeed, what I said had nothing to do with the SE generator: I have no idea which generator Urhegyi uses.

I was speaking of the most current top-down generators - which are indeed very biased (but less so than the bottom-up ones).
But even if you consider a really random and unbiased generator, the probability of drawing a puzzle not in T&E(1) is so low it can't be estimated.
In the ~6M puzzles of the controlled-bias collection (easier than an unbiased one), there's not a single one not in T&E(1).

The only thing I can say is, if Urhegy's generator finds a T&E(W2, 1) puzzle, it must be strongly biased - unless he has generated (and rated) hundreds of millions of puzzles.

.