hendrik_monard wrote:I am examining if there could be a role for SHC in my generating script(s). First, I am investigating the genealogy of my high BpB ranking puzzles, especially how the BpB evolves along the paths leading to these puzzles. On the path to one of them I came across this fine specimen:
987......6..8...........5...4.32..51...1.42.8....5834...32.581.....134.........2. B13B
Denis, could there be a bug in SHC, or is it a genuine B13B?
Hi Hendrik
Congrats for this great finding.
I have no doubt it's really a B13B.
Of course, no non trivial software can ever be proven bug free. But see section 3.2 of [HCCS] for all the support there is for the correctness of SHC.
The BpB rating is computed by progressively increasing p in T&E(Bp, 1). Which implies that even for levels BpB that have few instances, the cases of lower levels count as validations for level p in T&E(Bp, 1), provided that the computations for B are valid. But the latter have already been widely tested wrt SudoRules.
In the present case (SER = 11.7), no direct proof showing that SudoRules provides the same answer B13B is possible. It would take days of computation (that's why the SHC was developed). But what I've done is run T&E(B5) and I get the same RS as SHC (it is for making this kind of comparisons possible that we have recently added the printing of the RS after each BpB level):
- Code: Select all
+----------------+----------------+----------------+
! 9 8 7 ! 5 34 12 ! 16 36 246 !
! 6 25 245 ! 8 347 12 ! 179 379 2479 !
! 13 123 124 ! 4679 679 679 ! 5 8 247 !
+----------------+----------------+----------------+
! 78 4 689 ! 3 2 679 ! 679 5 1 !
! 357 35 569 ! 1 679 4 ! 2 679 8 !
! 127 127 1269 ! 679 5 8 ! 3 4 679 !
+----------------+----------------+----------------+
! 47 679 3 ! 2 4679 5 ! 8 1 679 !
! 28 679 28 ! 679 1 3 ! 4 679 5 !
! 145 1679 15 ! 49 8 679 ! 679 2 3 !
+----------------+----------------+----------------+
After this point, SHC finds NO elimination in T&E(B6) .... T&E(B12).
The first elimination it finds is n7r7c5, which should therefore require T&E(B13); this is what I'm currently testing with SudoRules (using function try-to-eliminate). I'll update this post when the computations are finished. (Of course, it's only a very partial proof.)
If I was you, what I'd now do is scrutinise the whole neighbourhood of this puzzle and of the path that you found leading to it, and try directly to increase SHC.
This example reminds me of a puzzle by Mauricio, with W=31 (and B=19). It appeared very unexpectedly and very far from the largest known W. As of today, nobody knows if one can find still larger values of W or B, and/or close the gap with the known ones. At that time, there was no way of trying to maximise W/B directly; so it more or less fell into oblivion - except for finding a special place in [PBCS].
Regarding the B rating, I think if you find some puzzles in T&E(1) near the above ones, you should not discard them but try to maximise their B rating (still with SHC).
Final remark: this puzzle has a non-degenerate tridagon, as all the puzzles recently found with BpB >7. It has a single guardian - which tends to produce the hardest puzzles.
.