Author: Bob Durrett
Date: 17:24:43 01/15/04
Go up one level in this thread
On January 15, 2004 at 19:29:56, margolies,marc wrote: >It seems to me that if a time management function is properly implemented, then >it should not be an extrapolation of time spent. I agree 100%. One should not try to predict the opponent's future time usage in the game based on how much time the opponent took on the moves already played in that game. Just because the opponent moved slowly [or quickly] in the early part of the game does not imply that the same will happen during the remainder of the game. >Instead, a good time management >system must take the eval function of the position into consideration. Again, I agree. However, perhaps there are more things to take into account other than just "take the eval function of the position into consideration." One such "thing" is time remaining. Another is estimated number of moves that must be made in that time remaining. If the flag is about to fall, the engine had better hurry up! For example, in a 30 0 game, where each side gets thirty minutes for the entire game and no increment is added during the game, then when there remains only a few minutes, there is the problem of deciding how fast to play. If the game is in the early opening phase, it is reasonable to expect that at least twenty more moves would have to be played so that would calculate out to only a few seconds per move on average. This is a case where your idea of using the position evaluation function would have an application to time management. The position evaluation function would determine which phase the game is in. More could be done along these lines, especially in the endgame where endgame knowledge might be used to help estimate the number of moves remaining. Generally, one must estimate the number of moves remaining in the game [assuming no time increment is added at each move]. That is where statistics may help. Having a non-zero time increment merely complicates the estimation process, IMHO. In speed chess games such as 30 0, it is often practical to just make sure you have more time remaining on your clock than your opponent does. If he plays slowly you can too since his flag will fall first. >EG, if a >game is clearly winning one need only avoid blunders so in a zeitnot, the search >depth of the program might be abridged until time control is reached. "Abridgement" sounds a wee bit drastic to me. : ) Bob D. > > >On January 15, 2004 at 18:00:15, Bob Durrett wrote: <snip>
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.