Author: Robert Henry Durrett
Date: 09:26:49 06/10/02
Go up one level in this thread
On June 10, 2002 at 07:01:03, Russell Reagan wrote: >On June 10, 2002 at 06:41:07, Benny Antonsson wrote: >>Like this.... My stupid program wants to take the rook ! > >Since the position you posted isn't exactly legal since you have like 20 >something black pieces on the board, I tried this position: > >[D]4b1r1/4k3/1rp1p1p1/1pPpPpPp/pP1P1P1P/P2K4/3B4/8 w - - 0 1 > >Apparently Fritz isn't any smarter than your program, because it wants to take >the rook too! Fritz sees this as -4.71 in favor of black after cxb6, and it sees >the position as -8.84 in favor of black after the "next best move" Be3. I always >knew that deep down I was better than my computer! > >Russell For a USER of Fritz [I have been spending several hours per day with Fritz and CB8 for quite a few years now], this is an especially important issue!! For a USER, the practical problem is to find a way, in the near term, to live with this problem. Having faith in the chess programming community, I find it simplest to assume that this problem will magically go away eventually. But this is now, and the problem is here now. The practical problem for a USER is to find the correct way to interpret the analysis results provided by Fritz, or by any other chess engine the USER is relying on. This example illustrated very clearly that one cannot safely assume that the numerical value of the position evaluation is a **direct** indication of who's ahead and by how much. Theoretically drawn endgames, where one side may be up in material, also require special interpretation of the engine output. There are other types of positions where the numerical value of the position evaluation has to be interpreted correctly. Opening gambit positions are another example. Positionally locked pawn structures, however, really "bring home" [dramatically demonstrate] the huge difference between human "position evaluators" versus the current crop of silicon-based chess "position evaluators." The human evaluator sees that it is a draw within seconds, if not instantaneously. Why cannot the chess engines do that? Well, it's just a matter of programming. "The programmers" will get around to this soon enough. Faith, faith, faith. [Hopefully not "blind" faith.] Practical USER workaround: Become smarter at interpreting chess engine findings. Bob D.
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.