Author: Bob Durrett
Date: 16:33:15 11/04/02
Go up one level in this thread
On November 04, 2002 at 19:15:40, Joe T. Pangilinan wrote: > I have been using the Fritz Program mainly for game analysis. I noticed that >it's analysis starts from the last move - going forward to the first move. Now, >the idea of backward analysis would make sense, if the engine retains the >evaluation from previous move, so it would save time and not have to re-do the >previous evaluation. I know that the Fritz engine clears the hash table after >every move. > My question is, are there any engines that retains previous >evaluation/analysis after every move, and not clear it's hash table? Any >response would be appreciated. I, too, have felt irritation at the way Fritz seems to disregard evaluation from the previous move when it steps back to the next move for evaluation. However, after thinking about it a long time, I came to the conclusion that the way Fritz does it is the way I would want it done too. Suppose the "best line," and it's numerical evaluation, from a move evaluation were retained in memory and used as a reference. Then when Fritz stepped back to the preceeding move of the game to evaluate it, I thought the reference line would always appear to be the best line. But why should it? In the lines initially examined by the engine for the preceeding move of the game, the leaf nodes numerical evaluations might appear much better than the reference line. The reference line would be discarded. Perhaps, however, the reference line could be retained until the very end. Then, an instant before the computer makes it's next move, it could compare it's favored line to the reference line. Somehow, I expect the real chess programmers here to say that "the computer just doesn't work that way." So I live with Fritz like it is and "grin and bear it." What else could I do? Bob D.
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.