Author: Uri Blass
Date: 16:15:03 12/20/04
Go up one level in this thread
On December 20, 2004 at 18:59:00, Dann Corbit wrote: >On December 20, 2004 at 18:43:19, Uri Blass wrote: > >>On December 20, 2004 at 18:31:56, Dann Corbit wrote: >> >>>On December 20, 2004 at 18:27:37, Uri Blass wrote: >>> >>>>On December 20, 2004 at 18:22:24, Dann Corbit wrote: >>>> >>>>>On December 20, 2004 at 16:32:24, Jouni Uski wrote: >>>>> >>>>>>Shredder 8 reaches quite stunning averaga search dept. E.g. with only 15 >>>>>>second/move average with Pentium 2400 and 128MB hash it averages 15,2 ply! For >>>>>>comparison Fritz Bilbao and Gandalf 6 reach both same 12,7 ply. No wonder >>>>>>Shredder beats them 13 - 7 and 14 - 6. Of course 2,5 ply is a lot difference! >>>>>>Is there any amateur or pro engine, which can equal Shredder in depth - Junior's >>>>>>half plys can be forgotten. >>>>> >>>>>I can write an evaluation function that only counts the wood. It will beat >>>>>Shredder's depth, but Shredder will easily kill it on a 1/10 as powerful machine >>>>>all the time. >>>> >>>>Only writing evaluation that only count wood will not be enough to beat >>>>shredder depth. >>> >>>A simple PVS with null move and wood counting only will get 19 plies in a few >>>seconds. Is that what you mean by pruning? >> >> >>In what position? >>I never had only wood counting but I guess it is dependent on the position and >>in tactical position when there are many threats to win material it may get >>smaller depthes. > >Maybe it is easier with bitboards. Note: This eval is really horrible. Don't >use it for anything other than entertainment purposes... > >eval = >(bitcount(white_queens) - bitcount(black_queens)) * 9 + >(bitcount(white_rooks) - bitcount(black_rooks)) * 5 + >(bitcount(white_bishops|white_knights) - bitcount(black_bishops|black_knights)) >* 3 + >(bitcount(white_pawns) - bitcount(black_pawns)); >only takes a few cycles. > >Your move ordering will have to be OK for the eval, and you will need a hash >table. >But I think you can see how cheap this eval is to run. I understood that the evaluation is cheap. In my program I also can have eval=(numqueens[0]-numqueens[1])*9+(numrooks[0]-numrooks[1])*5+... The speed of the evaluation is not important here because programs usually do not use most of their time for evaluation. The point is that null move may prune most of the moves because of no threat but I still doubt if it can outsearch shredder and there are positions when many moves are not pruned because they threat to win material. Uri
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.