Author: Vincent Diepeveen
Date: 02:29:28 06/30/99
Go up one level in this thread
On June 29, 1999 at 14:48:52, blass uri wrote: > >On June 29, 1999 at 09:57:24, Vincent Diepeveen wrote: > >>On June 29, 1999 at 09:23:23, Ernst A. Heinz wrote: >> >>>On June 29, 1999 at 08:39:30, Vincent Diepeveen wrote: >>> >>>>On June 29, 1999 at 07:10:59, Ernst A. Heinz wrote: >>>> >>>>>Oh yes, >>>>> >>>>>in view of Vincent's long story and his many excuses why "Diep" did not >>>>>fare too well again (no hardware excuses possible this time), I like to >>>>>add that meaningful *testing* is obviously a crucial part of program >>>>>development -- not only in computer chess. >>>>> >>>>>If people decide to enter a world championship with an untested program, >>>>>this is fine with me. But then, they should also stand by their decision >>>>>and accept the blame in case of severe failure instead of whining about >>>>>their immature and buggy code. It was their own independent decision to >>>>>employ it in the first place, wasn't it ... >>>>> >>>>>=Ernst= >>>> >>>>I'm nowhere whining Ernst, i'm just analyzing what went wrong after >>>>i got a question to do so. At least i can explain what went wrong >>>>in my program. Can you? >>> >>>It was easy to see what went wrong with your program -- a buggy and not >>>even deadlock-free parallel search tends to produce random numbers as >>>its overall result quite regularly ... >> >>Yeah i know you like to joke about it, >>but i wonder how most professional chessprogrammers are gonna >>improve the weakest chain of their programs without knowing shit >>from chess. > >They get ideas from people who understand something about chess. >I tell [programmer] my ideas about chess based on games that [program] >does mistakes >and I believe that he also heard ideas from GM's. >Uri I take your word for it. It definitely will not make a program weaker, but programmer must understand precisely what went wrong before being able to implement the knowledge. I have a draughtsprogram and am suffering from the same problem there. Cuz search isn't important in draughts (we outsearch in endgame most teams by >10 ply anyway and in middlegame/opening by 2-4 ply), so Marcel Monteba, although used to teaching others how to play draughts, must tell me, an absolute draughts beginner, how to implement something that's way above my understanding level. If Marcel succeeds in that, then still usual a pattern doesn't work, as in contradiction to human, a pattern must be complete; exceptions that obviously for Marcel don't work, are not implemented in the program. However those bigtime problems i don't see as i already have big efforts to understand the patterns anyway :) Now that may something about how i play draughts, but my draughtslevel is above the level of most programmers that make chessengines. Greetings, Vincent
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.