Author: Gerd Isenberg
Date: 02:44:25 02/23/06
Go up one level in this thread
On February 23, 2006 at 02:24:58, Uri Blass wrote: >On February 23, 2006 at 01:53:58, Gabor Szots wrote: > >>I am just curious whether it is so extremely hard to implement underpromotions >>in a program. Vasik Propaganda Rajlich put up a poll at his site, allegedly in >>order to "Help us make the products that you want". The people here _want >>underpromotions_ but, although promised several times, Rybka is still incapable >>of it. Now is it really important for him what the masses want but the task is >>simply too hard? >> >>Gábor > >The task is not too hard. >The main problem is implementing it in a way that is productive to playing >strength. > >Movei searches underpromotions but it is possible that I can do it few elo >stronger by not considering them because underpromotions are rare in chess and >the program spend time on analyzing stupid lines with underpromotions. > >It may be possible to solve the problem but I need to test if the solution is >really productive. > >A possible solution may be underpromotion reduction but the problem is that in >that case underpromotion may fail high because of the horizon effect. > >Uri Another idea for movegeneration (except at the root and except hashmove is already underpromotion) is to postpone the generation of underpromotions until one unmakes a queenpromotion. If the queenpromotion fails high we save some work. If the opponents best reply is a capture of the promoted queen and the promotion is no discovered check, we can also safely skip generating and making underpromos as well. Otherwise we can generate knight-promotion - and if queenpromotion returns draw score (possible stalemates) we can generate rook- and bishop promotions. Gerd
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.