Author: James T. Walker
Date: 18:17:47 04/15/02
Go up one level in this thread
On April 15, 2002 at 20:27:02, J. Wesley Cleveland wrote: >On April 15, 2002 at 11:30:36, James T. Walker wrote: > >>On April 14, 2002 at 12:38:46, Sune Larsson wrote: >> >>>On April 14, 2002 at 12:32:07, Roy Eassa wrote: >>> >>>>On April 14, 2002 at 09:02:37, Sune Larsson wrote: >>>> >>>>> >>>>> [D]8/1P3kP1/2P5/2K5/5b2/2B2b2/8/8 w - - 0 75 >>>>> >>>>> >>>>> This is from the game Jurasek 2305 - Piza 2290, 1997. >>>>> White can make no progress so the game was drawn here. >>>>> >>>>> Evals from some programs: >>>>> >>>>> Yace +- 1.96 1.Kb6 Ply 17 >>>>> Junior7 +- 2.12 1.Bd4 Ply 22 >>>>> ShredderP +- 2.61 1.Bf6 Ply 19 >>>>> CM Nextas +- 2.66 1.Kb6 Ply 16 >>>>> Fritz 7 +- 3.25 1.Ba1 Ply 17 >>>>> CT 14 +- 3.98 1.Kb6 Ply 19 >>>>> >>>> >>>> >>>>If the humans were only rated around 2300 and Fritz 7 is rated over 2650, then >>>>Fritz's evaluation of +3.25 (or Tiger's eval of +3.98) must be correct and the >>>>humans' evaluations of +0.00 (drawn) must be wrong. >>>> >>>>Surely weak humans of 2300 can't understand more about a position than mighty >>>>machines of 2650+. >>>> >>>>Right? >>> >>> >>> Right! ;-) >> >>I don't believe modern chess programs understand any positions. Sometimes their >>eval functions match what humans think is correct and they are said to >>"understand" this position. ?? Most chess programs are written to play chess >>and not correctly evaluate board positions. That is they are designed to play >>the best moves in literally any situation that arises. Of course none can do >>that perfectly. If I had written any of the above mentioned programs I would >>not concern myself with the "score". I would only be concerned with the program >>making the best move in that position. The score is simply a means to decide >>which move is the best. The fact that they don't seem to agree on what is the >>best move here may be the telling factor. Maybe there is no "best move". I >>doubt that any of the programs would lose this position from either side against >>a 2300 player. That is more important than spitting out the exact score for the >>position. (IMHO) >>Jim > >The problem is when computers encounter positions like this in the search. They >think it is won (or lost) and will steer toward (or avoid) it, possibly passing >up a win (or losing). That's a good point. The problem is... there are literally millions of positions like that. You may fix 1 and lose 3 in the process. Seems like everything in life/chess is a compromise. I think programmers try to get the best overall settings they can. There will always be positions like this. 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.