Author: Harald Lüßen
Date: 12:14:10 02/19/06
Go up one level in this thread
On February 19, 2006 at 13:59:36, Bahram Namjou wrote: >I asked this question before but with no comments so I post it again! > >when the ponder is on and chess engine thinks and evaluates a position, all of >the further evaluations are only based on one predicted move that considered to >be the opponent best next move however never this predicted move will change >during "thinking time" no matter this "thinking time" is 5 seconds or 50 >minutes...can anyone explain this why?...thanks, bn I know two ways of pondering. When an engine makes a move emg_move1 it is the first move in the principle variation: eng_move1 opp_move1 eng_move2 opp_move2 ... a) The engine now assumes the opponent will play opp_move1. Internally it makes the move and starts thinking (pondering) about its own answer, obviously beginning with eng_move2. pro: when the opponent really plays opp_move1 the engine can answer immediately or just use the time advantage and the hash Table and the so far calculated depth. con: when the opponent makes another move the search has to start from the begining. This is like 'no ponder' if there is not just a simple transposition. Off course it has to correct its internal move on the board. b) The engine starts thinking for the opponents move, beginning with opp_move1. There could be a better move opp_move1a. There is always a new principle variation and most important there are new entries in the hash table. pro: when the opponent makes the move opp_move1a the pv and the hash table are full with information to answer this move. If the opponent plays a move opp_move1b chances are good that it was considered too and there are some infos in the hash table. With this strategy a huge thinking time can be used better. con: If the opponent plays move opp_move1 the hash table entries with the information belonging to this move may have been overwritten and we lost a half move of depth. There are engines for both strategies. Many experts say strategy a) is better. At least for good engines which predict the right move anyway. Harald
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.