Author: Dave Gomboc
Date: 17:40:45 09/06/02
Go up one level in this thread
On September 06, 2002 at 19:41:06, Arturo Ochoa wrote: >>I said that I guess that it can find the correct move in 90% of the cases. >>It does not mean that it is going to play correct moves in the opening in 90% of >>the games. >> > >Yes, it can play correctly the main lines of the Dutch, Budapest, Wolga, Queens >Gambit, QGA, Ruy Lopez, Italian, French, Sicilian Paulsen and about 450 >openings. > >You don´t know anything about openings theory. > >What I Conclude about your declarations. > >Your declarations corresponds to a beginner about 1100 Elo. > > >> >>I know something about opening theory. >>I am not expert in this field but I have stable rating that is close to 2000 in >>the last years. >> > >Considering the fact that you are recomending 1. e4 e5 2. De2 is a fact about >your level: A beginner that speak without any sense. > >Yes, where because you rate is only about 1100 and if you are a correspondence >chess player, of course, this is not your real elo. > > > >>I can learn the reasons for mistakes. >>I found that one of the reason for movei's mistakes is underestimating mobility >>so I increased my mobility evaluation and the version that is playing in the 3th >>divsion has bigger evaluation of mobility relative to the versions that were >>playing in the 3th divsion and the 4th division. >> >>I can look at a lot of mistakes and try to generalize some things to teach my >>program. >> >>I have also ideas that are not used in movei and I stopped the work on movei >>because I prefer to learn C better so I can check for bugs better in the future. >> >> >> >>> >>>There many good lines that are too complex of solved so the resultan position is >>>the essence of the thing. The resultant position should be a position that the >>>chess sillicon knows to manage. >> >>I do not believe that they are too complex to solve. > >Yes, tell this to Kasparov about what he is thinking. :))) This is your best >joke of the day. > > >No, you are messing all the time. Please, Uri, read a beginners book of opening >theory and we can debate with some sense. > > >>I believe that there are positions that chess programs do not know to solve >>because the programmers did not teach them the right rules of evaluation and >>extensions in order to solve them. >> > >Yes? Then, all the programs are wrong! > >>I believe that seeing a lot of examples can help to define small number of rules >>that are going to help to solve many of them. >> > >Maybe, you can believe in genetic algorithms but it is another thing that you >can prove any concrete thing. > >It is useless to justify throug your Engine (off topic of course) because it >would be the world champion and I haven´t seen it in any Tournament with good >results. > >What you are testing in your chess program has proved any conclusive thing? I >don´t see anything useful that you have showed in this long thread, by the way, >very funny. :))) > >It is a mess thread about how you don´t understand a coin about chess openings >thoery. > >Regards, Arturo. I think that you're completely missing what Uri is saying. One reason machines (and humans) learn opening theory is to save up time for thinking later, but that's not the only reason. Humans also learn by example. By going over good opening moves, they can learn to imitate them even in positions where there is no direct opening theory. However, standard chess programs do not have this capability. Sometimes chess engines get into difficulty in the opening. Often, the software developer will blame it on the book: how there was a typographical error in the book, or how the book didn't cover that line because its author didn't know about that cheap trap, and so on. But that's all just bullshit. Uri's point is (if I may anthropomorphize) that the chess engine ultimately has to take responsibility for the moves that the program plays. A chess engine abrogates this responsibility at its own risk. If it places 100% reliance on its book, and isn't capable of finding good moves in opening positions on its own, then it risks playing stupidly and getting its butt kicked every time another program gets it out of book early. If it can't verify that a move is sound and not a typo, it will lose games. That verification can happen before the game or during it, but it needs to be done at some point. You can see this in the evolution of machine-"booking up". It used to be that you just hand-coded a book and blip, blip, blip, out came the moves. But nowadays programs consider the frequency that moves have been played in the past, the prior results of games with those moves, they can elect to perform a search to verify on its own the quality of a move, and so forth. In games like checkers where computers are simply better players than humans, their own evaluation is often more important than what a human thinks, and they'll only decide to go with the move humans play if they seem to be worth close to the same value to the machine. Uri realises that he can discover weaknesses in his chess engine by examining the moves it plays. If some of those moves are opening moves, what's wrong with that? Not playing with an opening book gives him more chances to see his chess engine go wrong, and consequently more opportunities to improve it. Dave
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.