Author: Christophe Theron
Date: 10:12:32 12/05/99
Go up one level in this thread
On November 28, 1999 at 20:40:45, Robert Hyatt wrote: >On November 28, 1999 at 10:42:29, pete wrote: > >><snip> >>>Then that begs the question: how is he 'testing'? He isn't playing in any >>>public human events on any significant number of games. If you don't enter a >>>lot of human tournaments, and you don't play against computers, then what do >>>you test against? Certainly _not_ against yourself... That won't fly... >>> >>>However, in my case, I really don't give a swat... I have more than enough to >>>work on.. :) >>> >><snip> >> >>Christophe Theron has posted more than once that he tests his program against >>human players of his country letting Tiger play on very slow hardware to improve >>it and that he doesn't own any modern chessprogram at all and only sometimes >>tests autoplay against fritz2 and genius5 to see if the autoplayer code still >>works . >> >>If you are aware of this you are in fact saying he didn't tell the truth but >>tests secretely against the top engines to find glitches in them ; this is no >>nice style if you have no proof. > > >I don't have a clue about what you are talking about. I didn't say or imply >anything. I _asked_ a question... I play manual games against a lot of different players. The goal is not to "tune" against specific players. This word implies that I would try to find weaknesses in my opponent's play, and try to focus on this. It's nonsense. It would lead to nothing. I would win a game here and there because by luck the opponent's weakness would show up, but loose many many games because of my own weaknesses. Bob, as a chess programmer, you know that the real work is to get rid of the weaknesses of your program. So I use the manual games I play to find "holes" in Tiger's evaluation. Whatever your opponent is, human or computer, if he finds a weakness in your program that's interesting. So I watch manual games, locate the weaknesses in MY program, then write them down in a list. After many games, when a pattern appears in my mind, I try to fix it. I cannot even remember if it was a human or a computer that showed me a weakness. Then, when I have fixed something, I have a very long automatic procedure to test if the fix has broken something in my program. Because the main problem is that often what you consider as an improvement actually HURTS your program. This automatic procedure involves no commercial or amateur opponent. I have 4 computers to achieve this task, but they are not connected to each other. No auto232 involved at all. After several hours I get an answer, so I can either keep the change or discard it. My opinion about Auto232 matches is that there is too much randomness in them, mainly because the book lines are chosen randomly. And because these match test the book also. I just want to test my engine, I don't care about the book in this test. Because of this randomness, you need to play a lot of Auto232 games to get a reliable result. Auto232 matches is not an option in my opinion. Too much time to get a result, too much noise in the result. I think that chess programmers that use Auto232 matches are loosing their time and energy. It is possible to get better accuracy in much less time. Christophe
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.