Author: chandler yergin
Date: 15:08:17 11/10/04
Go up one level in this thread
On November 10, 2004 at 12:36:00, Anthony Cozzie wrote: >On November 10, 2004 at 00:48:26, chandler yergin wrote: > >>On November 10, 2004 at 00:40:07, chandler yergin wrote: >> >>>On November 09, 2004 at 17:10:02, Anthony Cozzie wrote: >>> >>>>You are missing the point, Chandler. >>>> >>>>You probably played through the game with Shredder 8 on. >>> >>> >>>I did NOT PLay through the game! >>>I started on Black's move 10, turned on 'infinite analysis' >>>3 lines. It found 11.a3 and 11.h5 almost immediatly. >>>I decreased the 3rd line, and let the other two run. >>>I then 'clipped' the analysis & Posted it. >>>There was NO learning! >>> >>>So your assumption, along with Uli's and Ed's are incorrect! >>>I can't make it any simpler. >>>CY >>> >>>I did analyze the game later, and saved it. >>> >>> >>> At that point, it >>>>found that some of the positions it originally thought were good for black, were >>>>in fact good for white, and it recorded them. You then returned to the initial >>>>position, and boom! it has learned from its mistakes. >>>> >>>>What is the difference you ask? The point is that Shredder would not play those >>>>moves on its own, without help from a human, without a very long time to search. >>>> >>>>anthony >>> >>>Try it yourself... >>>Don't lock on one variation, >>>Use 3 lines, start any place in the game.. >>>Watch when you get to move 11. >>>Clip analysis and please Post it. >>>Thanks, >>>CY >> >>Here's what it looks like: >>Ed - Rebel 12 >> [D]rnbq1rk1/pp1n1p1p/4p1p1/3pP1N1/1b1p1P1P/2NB4/PPP3P1/R1BQK2R w KQ - 0 1 >> >>Analysis by Hiarcs 8: >> >>1. ยต (-1.04): 11.a3 Ba5 12.b4 dxc3 13.h5 Bb6 14.Nxh7 Nxe5 15.fxe5 >>2. -+ (-1.85): 11.h5 dxc3 12.b3 Be7 13.hxg6 fxg6 14.Bxg6 Bxg5 15.Bxh7+ Kg7 >>16.fxg5 Nxe5 17.Qh5 >>3. -+ (-3.43): 11.Bxg6 fxg6 12.Qxd4 Nc6 13.Qd2 Qe8 14.a3 Ba5 15.b4 Bb6 16.Bb2 >> >>(Yergin, MyTown 09.11.2004) >> >>Note, I used a different Engine, so there was NO learning! >>I could have used 5 or 6 lines, and more Candidate moves would be evaluated. >>If you limit it to one Variation; that's all you will see for hours and hours. >>CY > >By that logic, I could put zappa in 50-best line mode and it would solve every >test position ;) > >anthony Nonsense! Look at the results of Computers in 'solving' Test Positions! Programs have Bugs! Relying on Analysis Modules is courting Disaster! They are a 'tool', nothing more, they have no intelligence, no strategy, they store and retrieve information. Their evaluations when 'out of book' are based on 'static' considerations in a quiescent position with no forced lines or combos, as programmed , weights given for King Safety, mobilty, pawn structure etc. etc.. until you get into EGTB's, then play is exact. Hey, I give great credit to the programmers of the Top Programs, they are Monsters! But... ;) CY
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.