Author: Robert Hyatt
Date: 14:32:08 02/19/06
Go up one level in this thread
On February 19, 2006 at 16:14:59, Bahram Namjou wrote: >>Thanks Mike...so why this feature has not been implemented in new engines? so >>called "multiple permannet brain". I thought that in pondering, the engine >>could evaluate similar to "infinite analysis" mode in which the best move >>gradually change according to depth of play ...Right now because mostly the >>predicted move is not correct, the pondering is not as efficient as we >>expect...regards, Bahram > >Here is your mistaeke. > >In most cases the pondering move is correct. > >I agree that in case that the opponent use long time then after enough time of >pondering it is better to ponder about something else but it will be a mistake >to spend months of programming for better pondering that may do the engine less >than 5 elo better when it is possible to get bigger improvement by better search >or better evaluation. > >Uri > >Thanks uri...you must be right and obviously depends on the engine but usually >in slow games the predicted move are not as accurate as I thought even in those >engines that keep the hashtables between moves...regards, bn I'd bet the hit rate is well over 50%. Crafty actually counts this, and for consecutive ICC blitz computer games I saw these numbers: 54 right out of 107, 36 out of 61, 60 out of 101, 27 out of 41, 23 out of 33, 30 out of 45, 38 out of 51, etc...
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.