Author: Uri Blass
Date: 04:46:36 09/25/00
Go up one level in this thread
On September 25, 2000 at 06:45:34, Ricardo Gibert wrote: >Which of the following is the most effective for a chess playing computer >program: >(1) Never consider underpromotions. >(2) Allow consideration of underpromotions. >(3) Only consider underpromotions if promoting to a Queen leads to a stalemate >or if underpromoting to a Knight gives check. I think that 2 is the most effective and is used by all the top program that I know except old versions of Junior. 1 can cause stupid mistakes I agree that 3 solves most of the practical problems(there are some rare cases when underpromotion to a knight without check is the only way to avoid mate or has an important threat that is not check) but there is almost no difference in speed from considering underpromotions. 3 can probably be used in the folloeing way: 1)consider only promotion to a queen and underpromotions to a knight if the underpromotion is with check. 2)If you see a score of 0.00 then start another search when you consider all underpromotions. 0.00 can be not because of stalemate but I think that most programs cannot know if underpromotion is leading to stalemate because they do not evaluate stalemates when there is a stalemate in the board. They evaluate stalemate only when they arrive to positions with no king and have to decide if the reason is checkmate or stalemate. Another problem is that stalemate can be a result of a combination so evaluating if promotion to a queen leads to immidiate stalemate is not enough(using the score is also not enough because the score can be positive because a move that is not promotion when underpromotion is the best move). Uri
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.