Author: James T. Walker
Date: 12:35:17 01/19/06
Go up one level in this thread
On January 19, 2006 at 10:05:37, Albert Silver wrote: >On January 19, 2006 at 09:49:21, Rolf Tueschen wrote: > >>On January 19, 2006 at 09:41:39, Albert Silver wrote: >> >>>On January 19, 2006 at 09:01:07, Rolf Tueschen wrote: >>> >>>>On January 19, 2006 at 08:45:23, Albert Silver wrote: >>>> >>>>>>I think that if a program doesn't know not to repeat a losing line >>>>>>then it deserves it's fate. >>>>> >>>>>There's only one problem: it isn't the engine's fault. >>>>> >>>> >>>>You mean it's not the fault of the engine if it makes a decisive fault in a >>>>chess game? I think this is a simple wordplay but it has no real value as a >>>>verdict. >>> >>>The choice of openings is not made by the engine, but by the book used and the >>>parameters chosen in the program (such as the Fritz interface, Arena, etc.). The >>>engine makes no opening choice of any kind. So this huge repetition of openings >>>is not Rybka's choice/fault but the fault of the openings settings used. Rybka >>>just played the openings it was told to. It had no choice in the matter when it >>>was told to play the same opening 31 times. >>> >>> Albert >> >> >>All agreed but why did it play the losing line if it's otherware is so strong? >>We always touch the same point as last year, as every year in CC. I always said >>that the pretended strength is a myth. These engines are not strong like our >>super GM. > > >First of all, being strong doesn't mean it is unbeatable. GMs win and lose, and >even the World Champion. The argument as to comparing a top engine with a top GM >is old and pretty pointless. The engines perform at a 2600+ level, but somehow >this makes people think that since the 2600+ are all GMs and we know what to >expect from a GM, the engines must also play like them. They don't. They have >their strengths and weaknesses, and when matched against GMs they perform 2600+. >The arguments on GMs getting tired, or only losing because of a tactical mistake >are pointless, and sound like the whining I hear at my chess club when a player >who was 'completely winning' loses a game. > >As to the opening, my objection (and you've changed the subject here) was that >the results were shown as a comparison of strength between H10 with special >settings against Rybka, and in the circumstances presented, I don't think one >could draw any such conclusions. > >What if I were to instead choose an opening where Hiarcs 10 lost badly to Rybka >and had them play THAT opening 31 times instead? Would this then be proof that >Rybka was far superior? No, of course not. > > Albert To my thinking this result simply shows the Beta has no "book learning" or "positinal learning" at this time. Seems like a waste of time to prove a Beta is broken beyond the initial find. I once caught Fritz with a learning bug. It was playing Junior and Junior won 26 consecutive games as black with the same opening. I notified Chessbase and they had Fritz fixed in very short time. One of the few quick fixes I've seen them come up with. (Although it was probably the Fritz programmer who fixed it) Jim
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.