Author: Russell Reagan
Date: 21:36:39 11/25/02
Go up one level in this thread
On November 25, 2002 at 21:28:32, Robert Hyatt wrote: >The concept of a "captures only" q-search is so flawed that tossing out a >few extra captures, or keeping a few unnecessary captures is basically >unimportant... There are so many positions where a capture is not the best >move that to hope for much from the q-search is mostly wishful thinking... Exactly. Let's remember, q-search stands for "quiescence search", which means "searching for a quite or peaceful position." So we're looking for a position that is peaceful and unlikely to change drastically so we can evaluate it more accurately. Clearly captures are "violent" moves, so we need to check them when searching for a peaceful position. However, they are not the most violent moves. Capturing a pawn will make the position change ~1 point. That's a little bit violent. Forking the king and queen is ultra violent, but it's not a capture. Rattling off a series of 10 checks in a row leading to mate is violent, even though there may not be a single capture. Checks and other tactics (which aren't exactly easy to hunt down) are probably more violent most of the time than your average capture. Most of the "violent" captures (captures that will change the score of the position drastically) are bad ones anyway, so as Bob says, a capture only q-search has some pretty big holes in it. Instead of thinking "generate captures", think "generate forcing moves". Captures, checks, some attacks...whatever forces the opponent or yourself to have to do something (or suffer the consequences). Then try to figure out a way to keep your q-search from exploding :) Russell
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.