Author: Bruce Moreland
Date: 08:09:56 08/05/01
Go up one level in this thread
On August 05, 2001 at 10:25:46, Pham Minh Tri wrote: >Hi all, >Some my questions as I start my EGTBs: >1) What have I to do when there are more pieces than I need (for probing EGTB)? >For example, I have table of KBNK, but an endgame is KBNPK. Give the "redundant" >pieces for opponent? Maybe you can use them in some circumstances if there are more pieces, but they are not designed for that. >2) I see that index for EGTB is not simple, so why do we not apply the same >technique as opening book: use hashkey? I guess one of the reasons is to save >space for TBs - use halves of them. But I think we could reverse the board, >re-calculate the new hashkey to match the TBs again? Or do I miss something? The reason you use a hash key when doing the opening book is that it's hard to generate a unique number that you can use to find the position. With EGTB, this unique number is easily obtained, so it becomes a matter of: here is a big pile of data, here is a number, go get the corresponding item. The problem is that the pile of crud is a huge pile of crud. If Nalimov didn't care that his tables were larger, he could just make it a big array (in a disk file), and just go grab the right one. He wants to make his stuff smaller than this, so the complication is due to the overhead of decompression and digging through an index. The hash key method would be much more implementation-dependent. If I were using Nalimov tables, I wouldn't want to have to use the same hash values that Crafty uses, and hash things in *exactly* the same way. bruce >Many thanks for any help. >Pham
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.