Author: David Rasmussen
Date: 12:27:31 05/31/01
Go up one level in this thread
On May 31, 2001 at 10:31:12, Gian-Carlo Pascutto wrote: >On May 31, 2001 at 09:38:33, David Rasmussen wrote: > >>On May 31, 2001 at 08:56:33, Gian-Carlo Pascutto wrote: >> >>> >>>I use the CB approach after you once described it here or on rgcc and >>>it's saved my ass a dozen times... >>> >> >>What _is_ the Cray Blitz approach? > >Instead switching the window to +inf, switch it to +150 or something >like that, and if it fails again +350, next +1000, and only then >to +inf. > >Works a bit like a poor mans-MTD(n,f) :) > I see. I did that once, and it was OK, but I took it out as I wanted to do it more systematically. I don't like magic numbers :) So I want some systematic way of guessing how much to relax the window. Similarly, I've been thinking about choosing my aspiration windows size at the root by the behavior of the root score. If root scores during iterative deepening is very constant, I wan't to have a small window. If it fluctuates, I want to have a larger window. Maybe the derivative of the rootscore could be used for something. Maybe something else.
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.