Author: Andrew Dados
Date: 16:10:50 09/19/98
Go up one level in this thread
On September 19, 1998 at 18:57:30, Inmann Werner wrote: >Something "nice" happened to Inmichess. > >By putting everything I heard from in, I also put some bugs into my program. >Although it is fast, has a nice evaluation and all kind of stuff, it played >sometimes really strange. In Tschoops summer Tournament, it got 4 points in >eleven rounds. > >I looked to change the evaluation function to play better, most the pawn >evaluation... but nothing, really nothing changed. I got angry and set some >variables incredible high and ... nothing happened. Some Nights later, I found >one variable set to 1 instead to zero. This variable disabled all (really all!) >pawn evaluation in the middle game. > >The really strange thing is, that Inmichess with this bug did not really play >bad. There were some nice games and I, on the first look, never thought, that >there is no pawn evaluation at all. > >When I fixed the bug, I thought, Inmichess now will play "much" stronger. But >the effect was about 100 Elo at all. (estimated by auto232 games) > >Maybe evaluation is not so important, as I thought (since the bugfix, my program >is much slower), or the pawn evaluation is silly (i dont think so) or there are >other strange bugs. It may sound strange, but the more evaluation factors (knowledge) you add into program it gets much 'slower' then extra time needed to execute more code... what happens is that now 'granularity' of positional values changes and to find best move (or just cutoff) program usually has to search MORE moves at given ply... At least that is my experience... - Regards - Andrew >A small "computer programmers brain damage report" from > >Werner
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.