Author: Murray
Date: 02:38:09 12/04/02
Go up one level in this thread
On December 02, 2002 at 18:01:00, Bob Durrett wrote: >(snip) > >(1) If other programs offer this same feature, how do these chess engines >produce the alternate moves and lines? Differently from the way the first move >and it's line is produced? > It depends on the technique used, there is more than one way to skin a cat. >(2) If a chess-playing program were optimized to display multiple lines, would >the search algorithms and position evaluation code be any different from those >of a chess engine optimized for one move [and it's line] only? The technique I use is to keep the alpha window open by a certain amount at the root position. As you know with alpha beta algorithms, when an improvement to the best move is found, alpha is increased to match the new best score, as we are not usually interested in inferior moves. However in this case we are interested in inferior moves, so if AT THE ROOT ONLY (depth = 0) we keep alpha as (best score - some window size) we get information about all moves that fall within the window, and in an efficient manner. >(3) If an engine is not optimized for this application, how much will the engine >performance be handicapped if it is forced to display alternate moves and their >lines? That depends on the size of alpha window I mentioned above. The window size selected will depend on how many alternative moves you want to see. For example to see all moves, you will need an infinite window and performance will be much worse. For small windows, showing all moves slightly inferior to the best, overall performance is worse but not magnitudes worse. If anyone knows a better way I'd be interested to hear. Murray
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.