Author: Maurizio De Leo
Date: 13:43:27 02/09/06
Go up one level in this thread
You could just make a normal search (without TB) from the root position and choose the highest scoring move from the ones TB indicate as a draw. On February 09, 2006 at 08:13:27, h.g.muller wrote: >You hit upon an interesting pronblem of TB play: for positions in the draw >sector the program hasn't the slightest idea what to play (except avoiding >losing moves). Where for a human there are draws that you are happy with to >salvage and draws that you couldn't win by a hair's breadth, the TB simply does >not make that distinction. > >So a TB-driven engine in a drawn KbppKb game with unlike bishops would see >nothing strange in sacrificing one pawn on the first move, a second pawn on the >second move, and its bishop on the third move. KKb is still a draw, so why the >fuss? Of course in KKb there is absolutely no hope left for a swindle, while if >the opponent did not have the KbppKb TB he might have conceivably made a >mistake... > >This problem is not so easy to solve, because to maximize the probablilty to >seduce the opponent to a losing error requires knowledge on which type of errors >he is likely to make. The obvious method of keeping the number of non-losing >moves as small as possible would only work against an opponent that plays the >moves randomly. In practice it leads to quick conversion, because if I capture >his Rook in, say, KrpKr, only the single move that recaptures the Rook is >non-losing. But in practice it is of course inconceivable he would refrain from >recapturing the Rook, so the result is certain conversion to a drawn KpK >end-game that is so trivial that he can not possibly bungle it, even if he does >not have that TB...
This page took 0.01 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.