Author: Robert Hyatt
Date: 10:36:51 09/07/01
Go up one level in this thread
On September 07, 2001 at 02:14:17, Bruce Moreland wrote: >On September 06, 2001 at 13:58:00, Uri Blass wrote: > >>Do you assume that all programs use the same extensions? >> >>I can imagine that programs with more extensions can see it in less plies even >>with perfect move ordering. > >If you write a program that has a functional hash table, no null-move pruning, a >sensible eval, and so on, it will solve this problem in no time at between 18 >and 26 plies, which is approximately what Bob said. > >What is happening here is that Tiger is doing very badly. I am pretty good at >screwing up Fine 70. Assuming that Tiger's hashing system is not simply broken, >my bet is that he's doing null-move in some tricky way in the king + pawn ending >and/or he's doing some sort of reduced depth search that's eating his hash >table. > >It's hard to know what anyone would extend here. The problem is not really >vulnerable to extensions. It's a coordinate squares problem. The kings move >around and black is zugzwang'd and white simply wins the f5 pawn. There are no >checks, no recaptures, etc. > >I've never found anything that works better here than an implementation of >alpha-beta that could be taken straight out of Knuth. With a transposition hash >table, of course. > >bruce I don't think it is extensions. I think it is a forward-pruning problem. When you store results learned after forward pruning, you can wreck things for many plies to come...
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.