Author: Bernd Nürnberger
Date: 04:57:27 04/01/04
Go up one level in this thread
On March 31, 2004 at 12:35:21, Renze Steenhuisen wrote: >>(1) >>I recognized that a different move ordering delivers very alternating >>results in my alpha beta search. (I implemented null move, pv search, >>mvv/lva, some history moves, transposition table, quiescence search ...) >>Is this a bug or is this a normal behaviour? >>For example, with 3 history moves (on initial position) i get: >>[...] >> 8. 0:01.12 993512 0.00 Nc3 Nf6 d4 d5 Nf3 Bf5 Ne5 Ne4 >>... and with just one more history move (4): >>[...] >> 8. 0:01.28 1088694 0.00 e4 Nc6 Nf3 e5 d4 Nxd4 Nxd4 h6 > >Seems reasonable. Mmmh, that still is not very reasonably to me to get much different results for different move orderings. The score of the moves itself should not be affected ?! Do you have an explanation? > >>(2) >>[move ordering: hash, caps, history] >I am doing almost the same (SEE instead of MVV/LVA, and using 2 Killers) and I >am getting the same numbers typically. I am ordering all remaining moves >according to their History value. But how can I get to 90+ percent that are typically for more advanced engines? Is the "secret" the iterative deepening which is suggested by another poster. Or maybe some sort of an excellent SEE procedure? I will do some experiments on that. > >>(3) >>[SEE: docs wanted] >Me too, can't find it... I am using sources :-) But what I really would like to >have is a good description of which techniques together cause a FH-% of 90-95% >And I can't find it... soo... I see we are both looking for the same thing :) Ok, will browse some sources to figure out more about SEE. Thanks for your answer. Greetings, Bernd
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.