Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Strange ChessTiger 2004 accessment

Author: Oreopoulos Kostas

Date: 15:09:12 11/12/04

Go up one level in this thread


On November 10, 2004 at 17:05:16, Harald Faber wrote:


>>>>[D] 1Q6/3b2pp/p1k1p3/8/4N3/Pp2b3/2q3PP/5R1K b - - 6 1
>>>>


>
>No. I know how Tiger evaluates. I often see draw positions which Tiger does not
>evaluate as 0.00, although Tiger shows 3-times repitition. For exact
>interpretation of Tigers evals I suggest to ask Christophe Theron.
>For me this looks like another case of "uh, my program shows +/- x.y but the
>position is not won/lost". The mistake is: You trust the evals too much. I
>*know* that even Shredder and Fritz sometimes show +2, +3 and do not win. So my
>suggestion is: Trust your own eyes, not (only) the engine's eval.
>
>You find examples within my Tiger 2004 tournament games on www.harald-faber.de


Harald. +2 means 2 pawns up. Based on score the computer makes decisisions.
In the position i show, there is no horizon effect, nor deep down the variation
a trick. In a 3fold repitition Tiger may show -0.16 but not -2.
THis is definately a wrong evalutation. When i play Kd5, or Qd3 it goes to 0.0
at once , so there must be an algorithm problem. (dont know terminology, smthg
like fail high..) but definately smthg is not ok.



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.