Author: Robert Hyatt
Date: 07:42:51 05/23/04
Go up one level in this thread
On May 23, 2004 at 02:26:18, Volker Böhm wrote: >On May 22, 2004 at 20:17:45, Robert Hyatt wrote: > >>On May 22, 2004 at 19:54:08, Volker Böhm wrote: >> >>>Hi, >>> >>>trying to understand why my IID does not work, I read crafty´s IID code. (I´ve >>>solved my problem, in crafty the solution is hidden in the "do_null" flag as it >>>is used as a "do_iid" flag too). >>> >>>But I have one question left: >>> >>>Crafty does IID only if Alpha == RootAlpha && Beta == RootBeta. >>>In IID Crafty calls "search" with current Alpha, Beta window and depth - 2. >>> >>>When IID search fails low crafty searches again with window -MATE, Beta. With >>>this new window crafty would not be able to "iterative deepening" as when >>>"search" is called with window (-MATE, Beta) he will not make IID most of the >>>times as -MATE is not a Rootwindow-Bound. >>> >>>For me it would make more sence to change the if statement to Alpha <= RootAlpha >>>&& Beta >= RootBeta. >>> >>>Am I missing something? >>> >>>Greetings Volker >> >>Yes. I only want to do a _couple_ of IID searches during a single normal >>search. Not hundreds or thousands. If the IID tst is done and either alpha or >>beta is not the same as at the root, no IID is done there... > >while searching alpha will only get larger, beta will only get smaller. Thus >alpha <= RootAlpha (or alpha <= -RootBeta when other side to move) nearly >allways get the same result as alpha == RootAlpha (or alpha <= -RootBeta when >other side to move). The only difference is in the rare case of a IID research >with (-MATE, Beta). >Thus I don´t think your answer fits to my question. I answered in the general case. Remember that _some_ do fail-soft, so alpha can be lowered or beta can be raised...
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.