Testing PROLOG Sudoku Solvers
* Used websudoku.com generated Five Sudoku Puzzles, Four Evil and One Hard.
* Puzzles results in Table Below:
| Puzzles | Original Code | Modified for Triples | Heuristic Sudoku |
| Puzzle 1 EVIL | Failed | Failed | PASSED |
| Puzzle 2 EVIL | Failed | PASSED | -- |
| Puzzle 3 HARD | PASSED | -- | -- |
| Puzzle 4 EVIL | Failed | Failed | PASSED |
| Puzzle 5 EVIL | PASSED | -- | -- |
* Original Code Testing:
One Evil Puzzle Passed and Three failed. Does appear to be triples issue
It is noted that for example Puzzle 1 there is a great deal of repetition of numbers in candidate lists
especially the numbers 2 and 4: There are 36 instances of 2 as a candidate and 23 instances of 4 as a candidate,
to a lesser extent there are 18 instances of 6 as a candidate.
* Modified Code Testing:
One of the Three Evil Puzzles passed when the code was modified for triples
If the code was modified correctly, then it could be ascertained that the triples code was a necessary rule for solving
that puzzle, however, the two remaining unsolved evil puzzles are still in need of rules to help them succeed.
* Heuristic Sudoku Testing:
The remaining unsolved Evil puzzles, Evil Puzzle #1 and Evil Puzzle #4, were solved using the Heuristic Sudoku code
The test file shows the output which demonstrates the guesses that the code is making. There is much less output for
Evil Puzzle #1 than for Evil Puzzle #4. One thing noted about Evil Puzzle #4 was there was absolutely no number 9 anywhere in the puzzle.
which I think made it a more challenging puzzle for the program. Evil Puzzle #1 had only one two
and then at least a couple representatives of each number. For thoroughness tested all previously passed puzzle with
the Heuristic Sudoku code.
---- Please note Success was determined by inputing Prolog's results back into WebSudoku.com -----