Author: Vasik Rajlich
Date: 06:17:07 01/17/06
Go up one level in this thread
On January 16, 2006 at 01:45:21, Keith Ian Price wrote: >On January 16, 2006 at 01:32:48, Uri Blass wrote: > >>On January 16, 2006 at 01:18:14, Keith Ian Price wrote: >> >>>On January 16, 2006 at 00:57:35, Uri Blass wrote: >>> >>>>The reason is simple. >>>>It is known that many programs simply evaluate repetition as a draw and rybka is >>>>one of them(many programs are going to see a draw in the same conditions). >>>> >>>>It is the simple solution to the problem of avoiding repetitions. >>>> >>>>programs of course can evaluate only 3 time repetition as a draw but it may >>>>cause them to search bigger trees for no good reason so it is not clear if it is >>>>better for playing strength. >>>> >>>>programs that evaluate first repetition not as a draw may waste time by allowing >>>>repetitions and in some cases it even may cause draw by the 50 move rule because >>>>in superior position they may not be able to be lucky to find the right plan on >>>>time when avoiding repetition may force them to find the right plan if they >>>>evaluate first repetition as a draw. >>>> >>>>I choose to evaluate first repetition as a draw because when I have 2 options >>>>when it is not clear which one is better I prefer the simpler option. >>>> >>>>Uri >>> >>>Well, I thought that might be the case, but why does the 32-bit version evaluate >>>it as losing, while the 64-bit version thinks it's a draw. And should the >>>program evaluate a non-forced repetition as a draw, if it is so far behind? >>> >>>kp >> >>I do not know Rybka 1.0 beta 32-bit see a draw when it get the game and not >>losing score for black. >>Maybe latest 32 bit version is different. >> >>Deep Shredder 9.02 x64 - Rybka 1.0 Beta, Blitz:16' 2005 >>b1r1qn1k/1rb3p1/2p1p2p/2N1P3/1P1P4/R2B1N2/3Q1PPP/2R3K1 b - - 0 1 >> >>Analysis by Rybka 1.0 Beta 32-bit: >> >>43...Rbb8 >> = (0.00) Depth: 3 00:00:00 >>43...Rbb8 >> = (0.00) Depth: 4 00:00:00 >>43...Rbb8 >> = (0.00) Depth: 5 00:00:00 >>43...Rbb8 >> = (0.00) Depth: 6 00:00:00 >>43...Rbb8 >> = (0.00) Depth: 7 00:00:00 >>43...Rbb8 >> = (0.00) Depth: 8 00:00:00 1kN >>43...Rbb8 >> = (0.00) Depth: 9 00:00:00 1kN >>43...Rbb8 >> = (0.00) Depth: 10 00:00:00 2kN >>43...Rbb8 >> = (0.00) Depth: 11 00:00:00 3kN >>43...Rbb8 >> = (0.00) Depth: 12 00:00:00 7kN >>43...Rbb8 >> = (0.00) Depth: 13 00:00:01 12kN >>43...Rbb8 >> = (0.00) Depth: 14 00:00:01 18kN >>43...Rbb8 >> = (0.00) Depth: 15 00:00:01 28kN >>43...Rbb8 >> = (0.00) Depth: 16 00:00:02 48kN >>43...Rbb8 >> = (0.00) Depth: 17 00:00:03 85kN >>43...Rbb8 >> = (0.00) Depth: 18 00:00:04 148kN >>43...Rbb8 >> = (0.00) Depth: 19 00:00:05 250kN >>43...Rbb8 >> = (0.00) Depth: 20 00:00:06 402kN >> >>(, 16.01.2006) >> >>Uri > >Ok. The difference between Chris' eval, and yours is that you loaded the whole >game, so that it could see the repetition. I still think with a score of 2.72, >it should not assume that the other side will take the repetition. It might have >found a better answer that would lower the score to 1.5 or so, but rejected it >for the 0.00 false score it came up with. > >thanks for the answers, >kp There are advantages and disadvantages to doing what Rybka does. I think overall it's a good idea, but it's hard to prove it (or to be sure). Vas
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.