Hi coloin
coloin wrote:I think there are already 6259 minimal puzzles in total in the 4 files.. in 781 different grids... so no losses in your scripts !!
4 files, with respectively 3362, 521, 634 and 1822 minimals = 6339 minimals (not 5339, my typo). Maybe you should check what you posted on google.
As for the rest, what I wrote remains unchanged.
BTW, this was just preparatory work, using none of my scripts; only gsf's software.
The main point is, we finally have the same numbers of minimals and solution grids. And this is my starting point for analysing the different T&E(2) layers for this collection. As they involve very large numbers of puzzles (much larger than in the T&E(3) case for collections of the same size), I needed some foolproof collection.
coloin wrote:denis_berthier wrote:The most surprising point is, most of those solutions grids (653/781 Edit: 776/781) are not in the 67359 ones of mith's collection of 4634101 minimals.
As I remember some new TE3 puzzles came from BxB6 puzzles as a by product of generating and minimizing BxB6 [ removing a clue maybe to find a BxB6+ or a TE3]
However no BxB6+ ever were found by adding a clue to a known TE3 puzzle ... as we see not surprisingly
Starting from a T&E(3) puzzle, or better its BRT-expand, one can only get an easier puzzle (by any of my classifs) by adding a candidate; in particular it can be at any T&E-depth.
If you then search for the minimals and keep only those in T&E(2), you can a priori get minimals with any BxB value.
It happens that you very rarely get a large BxB value that way. But there's no
a priori reason for this to be impossible; the question remains about the 5 solution grids I posted above, whether their BxB puzzles came before or after their T&E(3) ones.
What's quite surprising is, if instead of T&E(2) you keep only the puzzles in T&E(1), you will get lots of them with very high B ratings.
I currently have ~115,000 minimals in B12+ and a few hundred ones in B20+. See
http://forum.enjoysudoku.com/the-layered-structure-of-t-e-depth-d-t45647.htmlfor the detailed (simple) procedure I used to get them.
.
