I wonder if there are tablebases position when program never can see the right move(without tablebases) because of pruning by hash tables.
The point is that the same position that may be mate or draw by the 50 move rule in different cases and the program may memorize wrong information in the hash.
A possible test can be done by building tablebases based on the 5 move rule(instead the 50 move rule) and testing programs if they can find without tablebases by search the right move and the right result in all positions(of course you need to modify the 50 move rule and have instead of it the 5 move rule but it should be simply changing a constant in the program).
Note that in this case many KQ vs K or KR vs K are a draw.
50 move rule and maybe bug in chess programs
Moderators: hgm, Rebel, chrisw
-
- Posts: 10298
- Joined: Thu Mar 09, 2006 12:37 am
- Location: Tel-Aviv Israel