JPF wrote:Apart from this, I can see no reason why a pattern should generate puzzles only in the 7.x range.
This pattern:
- Code: Select all
+-----+-----+-----+
|x . x|. . .|x . x|
|. x .|. . .|. x .|
|x . x|. . .|x . x|
+-----+-----+-----+
|. . .|x . x|. . .|
|. . .|. x .|. . .|
|. . .|x . x|. . .|
+-----+-----+-----+
|x . x|. . .|x . x|
|. x .|. . .|. x .|
|x . x|. . .|x . x|
+-----+-----+-----+
contains only 7 non isomorphic valid puzzles, found by
blue.
All ratings are 7.2 x:
Yes, but that's purely factual. After browsing the reference, I can see no general reason why it should be so. It's just the result of systematic computations. Nice pattern and result, though.
Generally speaking, there are no criteria that'd allow anyone to tell in advance that all the puzzles for some pattern are in some given range.
JPF wrote:btw, it seems unwise to use gsf's program to establish the rating distribution of valid puzzles compatible with a given pattern.
We have to be sure that the generation of these puzzles is carried out without bias.
Personally, I am not aware of the methodology used by gsf.
I never mentioned the real distribution for urhregyi's pattern. I informally compared the puzzles generated for it with those for other patterns, using gsf's generator in all cases (i.e. with the same bias in all cases). I don't think all this is worth any more formal analysis and I don't need it to be intimately certain that the unbiased mean rating for puzzles with this pattern is much higher than the general unbiased mean rating.
