Author: Tord Romstad
Date: 07:00:26 01/25/04
Go up one level in this thread
On January 24, 2004 at 18:35:22, Uri Blass wrote: >On January 24, 2004 at 17:12:45, Tord Romstad wrote: > >>On January 24, 2004 at 15:57:50, Mike Siler wrote: >> >>>In an average middlegame position, around 80-85% of the nodes my program >>>searches are quiesce nodes. I have a static exchange evaluator and I only search >>>captures with SEE value > 0. It seems like other engines are always under 25% >>>qnodes. What else should I be doing to reduce these numbers? >> >>Use the SEE more aggressively. When the static eval is below beta, but >>static_eval+(value of capturing biggest hanging enemy piece) > beta+margin, >>return beta. This is too risky unless your SEE is very sophisticated. There >>are >>two ways to solve this problem: You can improve the accuracy of your SEE, but >>this tends to make it much slower (of course). You can also use your static >>evaluation function to estimate the tactical complexity of the position, and use >>this estimate to decide whether it is safe to trust your SEE at this node. If >>there >>are pinned, trapped or overloaded pieces or too many pieces are hanging, you >>search the captures, if not you just return beta. >> >>I use the second approach. My SEE is rather simple, and my qsearch uses >>information >>computed by the static eval to decide whether (and which) captures should be >>searched. >> >>Tord > >This post seems to contradict another post of you >http://www.talkchess.com/forums/1/message.html?343947 > >You said there that the problem with futility pruning is making assumption about >how much a move can change the evaluation and I read now that you do exactly >that(make assumptions about how much a capture can change the score). In that post, I was talking about futility pruning in the main search, which I don't do. I use futility pruning in the qsearch, though. Ideally I would prefer to avoid them there, too, but the cost is too big. So in this particular case, there isn't really a contradiction. However, I'm sure you can often find real contradictions between things I have posted at different times, too. What I do and don't do changes very rapidly. Tord
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.