Author: Dave Gomboc
Date: 14:55:45 06/11/99
Go up one level in this thread
On June 11, 1999 at 16:44:52, KarinsDad wrote: >On June 11, 1999 at 12:57:57, Dave Gomboc wrote: > >[snip] >>> >>>Point well taken. >>> >>>I guess variant chess is too strong of a term since Junior does allow the >>>opponent to underpromote. >>> >>>I guess (slightly) inferior chess is a better term. >> >>But, because it might compute quicker than if it dealt with underpromotions, it >>may actually play better. >> > >Doubtful. > >Generating the other promotions when a pawn promotes is no big deal. Since pawn >promotion is a relatively rare event in chess (even in the searches) it would >not lower the NPS enough to make a real difference (relatively speaking, there >could always be an exception). > >If you ordered the promotion such that queen promotions occur first, you would >often alpha or beta cut out the underpromotions (a queen promotion is a major >shift on the material part of the evaluation). It is only in special cases when >a queen promotion doesn't do much of anything (i.e. it does not change a lost >game to a draw or win; or it draws the game due to an immediate stalemate) where >the underpromotion would really be considered anyway. > >If a program is well written, this does not seem like a biggy and the program >should actually be weaker (1 or 2 ELO) if it cannot find the g8N+ type of moves. > >But, depending on how Junior is coded, it may have been more of a pain in the >butt to implement than the 1 or 2 ELO was worth (i.e. fight the big fires >first). > >KarinsDad :) Searching the underpromotion is so often a waste of time that I would expect an approach that does not search it to be better. At the top level, everyone is fighting for an elo here and there in the software. Generating the underpromotions does add some extra overhead too, which can be minimal or slightly significant, depending on how you are doing things to begin with. My assertion is testable. However, it is not so important to me that I intend to autoplay 50000 games to find out. :-) Dave
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.