Author: Vincent Diepeveen
Date: 02:58:11 08/12/99
Go up one level in this thread
On August 11, 1999 at 21:57:23, Bas Hamstra wrote: >This is not a good example of how SEE can error. However there are many examples >possible. I too think SEE can miss things, like all sorts of pins and throw out >captures when it shouldn't. However, as a percentage of all captures, how often >does that happen? I think there is consensus that the errors are more than >compensated by the reduction of qnodes. There is no consensus about this at all. I'm the opposite. It's like the real dumb program Bob and i are discussing about, in the end those stupid things will lose as tactics aren't the weakest chain in chessprograms. If you use SEE for SURE your eval is the weakest chain. You con't even see exchanges correctly, or you grab poisened pawns. A SEE cannot be combined with a big evaluation. A see works brilliant for piece square table programs though, as junior quite obviously shows, as they don't lose knowledge when applying it. >A cheap alternative is to throw captures out when after a capture the capturing >piece is hanging (and A > D). >Regards, >Bas Hamstra. > > > >On August 11, 1999 at 16:12:07, Tom King wrote: > >>I must be missing something obvious with using SEE to prune >>in the Q. search. If I understand right, "losing" captures >>(according to the SEE) are pruned right out. But I fail to >>see how this won't screw up big time. >> >>Let's take an example. Now assume we have a white bishop on G5 >>and a black knight on F6, and the knight is protected by a pawn on G7. >>Now the SEE might well assume that BxN is a loser (assuming that >>the bishop is worth a fraction more than the knight), because >>after BxN, gxB, white has lost a bishop for a knight. But it >>might be that this is in fact a very good capture, because it >>destroys black's kingside. And a program with the SEE Q. pruning >>might not want to play this? Am I missing something obvious >>here? >> >>Cheers >>Tom
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.