Problems with Analyzer

[crosspollinator]

I've just downloaded SudoCue and have been experimenting with the Analyzer and Optimizer on my own puzzles. I used Optimize with Keep Symmetry on this puzzle:

Code: Select all

...9...2.
..7.8.3..
.9...6..8
3..8...5.
..4.9.7..
.6...2..1
4..7...8.
..1.4.2..
.5...1...

and it gave me this puzzle:

Code: Select all

...9...2.
..7.8.3..
.9...6...
3..8...5.
..4.9.7..
.6...2..1
...7...8.
..1.4.2..
.5...1...

The Analyzer rated the original puzzle at 1124, and the Optimized puzzle at 903. According to the Analyzer, removing two of the starting numbers made the puzzle easier.

Another problem with the Analyzer is that if I open a new puzzle with the Analyzer window open, it adds the solving steps for the new puzzle to the steps for the old puzzle, and gives a score that is too high.

The version of SudoCue that I downloaded was 1.3.0.

[Ruud]

Hi, thanks for your very interesting observations.

I tried both puzzles and indeed, the non-minimal puzzle gave a higher rating.

I also found the reason for the difference.

The non-minimal puzzle allowed SudoCue to find more techniques to apply. In doing so, the score was incremented, but these techniques did not make a breakthrough, so the heavier scoring steps were still required. This problem has been discussed on other forums, but trying to ignore individual steps would make rating of puzzles a very complicated process that takes too much time.

For now, you'll have to live with this, but I know other programs have this problem too.

Another problem with the Analyzer is that if I open a new puzzle with the Analyzer window open, it adds the solving steps for the new puzzle to the steps for the old puzzle, and gives a score that is too high.

I cannot replicate this problem. Could you describe the exact steps? I tried paste, open from file, page up/down through a .sdm file, but all seem to work fine.

cheers, Ruud

[crosspollinator]

Thanks for the explanation of why the non-minimal puzzle had a higher score than the minimal puzzle. My own puzzle maker's rating algorithm doesn't do a very good job at all. Rating puzzles seems to be a harder problem than creating them.

The other Analyzer problem occurs when I open a puzzle and invoke the Analyzer without clicking Restart first. For example, if you save the first puzzle above in a file, then open it with SudoCue, and invoke the Analyzer, it gives a score of 2539. If you click Restart first, it gives a score of 1124. I'm running on Windows XP.

[crosspollinator]

In looking at the other Analyzer problem again, I think I got it backwards. I think that the Analyzer produces the correct results when the puzzle is first opened, and incorrect results if you click Restart, and then Analyze again. I think that the first puzzle's score should be 2539, not 1124.