18 | May | 2009 | brainwagon

Earlier, I blogged about the historic book available from Google Books called Gould’s The Game of Draughts. While surveying a few problems, I noticed that my play was significantly different, and included a number of different outcomes for a couple of tested positions. So, I decided to use the newly added 8 piece Cake endgame database (thanks Martin, for releasing the code and the databases) and see if there were any mistakes I could find. Mind you: these are mistakes which the database points out Immediately, it doesn’t actually point out mistakes in play that require search. I found 47 errors, which for fun, I’ll list below.

I think I am going to (for fun) systematically work through these, and document the improved lines.

Addendum: I seem to have a bug in my code. The list wasn’t correct. I’ll replace it when I verify the results.

Addendum²: Here’s my list of 28 mistakes. I think it’s right, but it might not be, I’m still confused by some of the database issues.

916 : Gould’s Problems #30 is actually a loss for white
986 : Gould’s Problems #97 is actually a draw
1019 : Gould’s Problems #130 is actually a loss for red
1023 : Gould’s Problems #134 is actually a loss for red
1025 : Gould’s Problems #136 is actually a draw
1033 : Gould’s Problems #144 is actually a draw
1086 : Gould’s Problems #197 is actually a loss for red
1094 : Gould’s Problems #205 is actually a draw
1107 : Gould’s Problems #218 is actually a draw
1108 : Gould’s Problems #219 is actually a draw
1171 : Gould’s Problems #281 is actually a draw
1404 : Gould’s Problems #516 is actually a loss for white
1658 : Gould’s Problems #767 is actually a loss for red
1660 : Gould’s Problems #769 is actually a draw
1668 : Gould’s Problems #777 is actually a win for red
1677 : Gould’s Problems #786 is actually a loss for white
1679 : Gould’s Problems #788 is actually a draw
1694 : Gould’s Problems #803 is actually a loss for red
1770 : Gould’s Problems #879 is actually a draw
1774 : Gould’s Problems #883 is actually a draw
1775 : Gould’s Problems #884 is actually a loss for white
1817 : Gould’s Problems #924 is actually a loss for red
1821 : Gould’s Problems #928 is actually a win for red
1902 : Gould’s Problems #1009 is actually a loss for white
1923 : Gould’s Problems #1030 is actually a win for red
1930 : Gould’s Problems #1037 is actually a draw
1958 : Gould’s Problems #1065 is actually a loss for white
1972 : Gould’s Problems #1079 is actually a loss for white

Addendum³: Check out this position from the list above with nine pieces:

The database code sees this as a draw, in spite of the fact that it only has up to eight pieces, because the capture resolution code removes one capture from the game before it does any database lookups.