Author: Albert Bertilsson
Date: 01:42:14 12/01/03
Go up one level in this thread
On December 01, 2003 at 03:23:06, Roberto Nerici wrote: >Apologies for the off topic post :-) > >I'm implementing some basic time handling code at the moment. I've found a few >bits on how to decide how long to think, some in threads here, some in some >books I've got. Something I haven't seen is how/whether to modify the allocated >time based on the number of moves available to choose from. > >It's obvious that if you only have one valid move you can make you just make >that move instantly. I've got my engine to handle that. However, what if there >are two moves to choose between? In a mid-game position where perhaps there is a >check involved, there might be only two available moves and one is much better >than the other. Does anybody's engine try to handle such a situation as a >special case? I've got my engine to compare the scores of the two moves and if >at the first check time (the point when my engine decides if it's searched >enough or should carry on) there is a large difference, it picks the higher >scoring move. However, I wonder if this is too risky, and of course there is the >question of how "large" the "large difference" should be. > >Thanks for any thoughts on the matter >Roberto/. > >PS: you can all now get back to long, tedious hardware threads, attacking the >ICGA, etc :-) I like your idea, and I've thought about it numerous times, because very often one move is clearly better, this is not specific to checks but the engine still spends lots of time to search the position and eventually pick that simple move. The problem is that the only way to guard yourself against tactical blunders is to search the position and see if there is anything dangerous. I guess that one could perhaps do someting like: if (one move is much better than the rest && depth > 5) reduce time by half I haven't done this because I think it is to dangerous, search is needed to avoid tactical problems, specially against other computers. This scheme could perhaps work better against humans, but I'm not sure. /Regards Albert
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.