wintder wrote:I did an 8 hour "rate" today and came up empty.
Here is my log about what I did in this last game:
First I was lucky enough to see this game when it was beginning, so I give it a try to the pattern, and my generator could output about 1 puzzle/second "randomly", that is very fast for random generation, but it is slow for "specialized" algorithms (n off/n on). I generated about 100 random puzzles and then input them to the specialized generator.
Having heard of wintder warning on ratings, I decided to add a rating constraint in my generator (a rating that I have implemented, more or less correlated with SE on the higher rated puzzles). As you can see, my first 3 puzzles had ratings 9+, and they were the only 3 I rated, they were the highest rated by my program.
As the next option was to mass-rate puzzles, then I decided to filter only the easiest puzzles that are not singles only (according to my rating). Having generated ~3000 of those puzzles in no more than 10 minutes, I rated them with SE, it took no more than 15 minutes. Long ago I coded something to analize the output given by SE, then I posted what I found.
Then I give one more try to my generator, and ran it until I had 10K puzzles, then I rated them with SE (50 minutes) and analized the output. I found just one more rating, then I decided to call it a day.
At night I decided to look exhaustively for all puzzles that are automorphic if you reflect along the antidiagonal (50 minutes). I found only 114 essentialy different puzzles of that kind, and just one more rating (9.8), in fact, take this puzzle:
- Code: Select all
+-------+-------+-------+
| 1 . . | . . . | . . 2 |
| . 3 . | . 4 . | . 5 . |
| . . . | 6 . 7 | . . . |
+-------+-------+-------+
| . . 2 | . . . | 8 . . |
| . 8 . | 2 . . | . 9 . |
| . . 5 | . 7 . | 6 . . |
+-------+-------+-------+
| . . . | 5 . 4 | . . . |
| . 6 . | . 9 . | . 1 . |
| 7 . . | . . . | . . 3 |
+-------+-------+-------+ ER 9.8
do c123456789->r987456321, and you get
- Code: Select all
+-------+-------+-------+
| 2 . . | . . . | . . 1 |
| . 5 . | . 4 . | . 3 . |
| . . . | 6 . 7 | . . . |
+-------+-------+-------+
| . . 8 | . . . | 2 . . |
| . 9 . | 2 . . | . 8 . |
| . . 6 | . 7 . | 5 . . |
+-------+-------+-------+
| . . . | 5 . 4 | . . . |
| . 1 . | . 9 . | . 6 . |
| 3 . . | . . . | . . 7 |
+-------+-------+-------+
if you reflect along the antidiagonal and do the digit permutation 1->1,3->3,9->9,2->7,7->2,4->8,8->4,5->6,6->5, then you obtain the same puzzle, using a uniqueness technique we asure r7c3=r3c7=9 and then the rating drops to 6.6. And that was it.