Author: Antonio Dieguez
Date: 17:22:09 11/05/99
Go up one level in this thread
On November 05, 1999 at 15:03:46, Bruce Moreland wrote: >On November 05, 1999 at 07:04:35, Antonio Dieguez wrote: > >>I just implemented recursive null-move, with a reduction of 2 plies. >> >>Obviusly a 1 ply reduction is safe enough, but 2, wich is the usual, for me >>looks not so safe, for example trusting that a yours-mine instead a >>mine-yours-mine-yours will give a minimum is a little optimistic isnt? and >>anybody have done a test about the percent of mistakes of the null-move or >>something? I mean when the score of the best move at the position, without >>null-move, results worse than the result doing null-move. >> >>Also am seeing not so much improvement, just from 1/2 to 1 ply. >> >>Another question, where do you make it works? in every node if eval>beta+a (a >>could be 0) from profundity let say n to the final or from a constant distance >>of the final (from final-n to the final) ? >> >>Thanks.Sorry my bad english and have a nice day! > >One ply can mean that you go two or three times faster. > >Do some tactical tests. I will do some tests tomorrow, but I am called "the lazy testerman". >I don't have any information about rate of error, but I know that R=2 seems to >find tactics faster than other settings, and way faster than not using null move >forward pruning at all. yep, it seems in some games that my program defeats more easily faile 0.6 >Sorry, but I didn't understand your final question. the final question was where do you use it, how much near the root or the leaves, could be great use it ever (if eval>beta) from for example 4 plies after the root, to the final, but it could miss something forever, or not? >bruce Thanks Bruce. Antonio.
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.