Author: Rolf Tueschen
Date: 05:26:43 06/17/04
Go up one level in this thread
On June 17, 2004 at 08:09:01, Vincent Diepeveen wrote: >To get hundreds of points more in WM test all i need to do is next at the end of >my evaluation: > > score -= (lightpiecescores>>1); > score += (passedpawnscore>>1)+kingsafety+mobility; > >Please tell me whether you understand what i mean above. > >I feel this is the fundamental problem you do not want to understand. > >In 99% of all games you win when you capture a piece. >In testests you solve the position by giving away a piece. > >It is the fundamental problem why testsets do not work. > >WMtest is the biggest patzer test in that respect. > >Look at Tiger1 versus Tiger2 score in endgame. > >Tiger1 was at the first release of the WM test the best endgame program >according to the test. > >What is the difference? > >Tiger1 just gives 3 pawns bonus for a passer at 6th row :) > >So very stupid and simple bluffing knowledge solves everything there. > >That's what the entire WM test is about. > >The proof has been delivered already by Tiger1. > Thanks for this clear statement that also a lay in programming could understand. Here is a question that still needs to be answered because it's the main argument from the defenders' side. Gurevich is never tired of that. He argues: his WM-Test does NOT calculate and result into numbers about "strength" of the engines but "analytical abilities (to solve positions with a clear best move)", he says the more of such positions a program can "solve" the stronger it is also in tournament chess. He says also that it is very telling that their rankking lists correspond so good with those ranking lists for strength. So, he concludes, his researches analytical ability (AF in short in German) is the base for strength overall. Now my direct question to you: where do you still see the disadvantage of such "AF" rankings and is it sound to claim NOT strength research but such analytical ability? Is it just a trick of argumentation or where is Gurevich somehow blinded? (If you could answer this with the same clarity, also in translatable German it would help in the whole debate. Just a last hint: In CSS I can read some supporters of the WM-Test and they all report that they hasd much fun in their own testings and that the results were quite ok. Why do _they_ feel ok, when _you_ say, as a programmer, _you_ can't exploit the results from such a test (with positions from human chess)?)
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.