Author: Alessio Iacovoni
Date: 01:55:20 08/28/98
Go up one level in this thread
On August 28, 1998 at 03:51:21, Serge Desmarais wrote: >On August 28, 1998 at 03:24:40, Alessio Iacovoni wrote: > > > >> >>Serge i'm not a programmer.. but i've studied some logic (philosophy) and i know >>something about sound reasoning... so.. it appears to me that a strategy that >>tries to develop "specialization" is going to be more efficient and successful >>than one using only one single approach. Now... tactical engines have gone very >>far.. they are capable of beating strong players, very strong players, sometimes >>internationa masters and grandmasters. The question that arises is "why is it >>that they cannot always beat a grandmaster"?. The answer: "grandmasters now how >>to exploit the weaknesses of the program. So.. what does this bring us up to. >>logical: to the need of a "positional blunder check" engine.... as dynamic as >>possible i.e. capable of "modifying" its strategy in order to prevent anybpdy >>from expoliting its weaknesses. So there is the idea.. it's not original i >>believe because i've just read an article on it on alt.computers.chess that >>talked about deep blue and "hsu"... but it is the only viable one... If THIS >>strategy was followed I guess a double processor would be suffcient to beat a >>grandmaster always. >> >>By the way... You now.. I'm Italian.. and italians are wonderfull at coming up >>with ideas... but then we leave the dirty work of implementing them (and getting >>the credit for them)to others.... :)))) > > > Well, the GMS don't win simply because they exploit weaknesses in the >program, but also because they understand the position better. That's what I meant. I remember a game >of Bronstein against a computer (VERY tactical!) and Bronstein sacrificed a hole >piece without making ANY calculations! He blasted the computer at tactics! Well.. I wouldnt call that tactics but strategy. The very fact he didnt do any calculations means that it was a strategical idea. After >the game, he just said that "by experience", he KNEW the sacrifice was good and >would probably win. > That's exactly how to win a program.. rely on experience and strategy and not on tactics. > Another example of that: I was examining a game of Tal vs Smyslov (or was it >Smyslov-Tal?) with Genius, on my Cyrix P150+. Tal gave a hole piece and his >pieces had a tremendous activities. Genius was saying that all the tactics of >that tricky position were winning for Smyslov, despite the comments of Tal in >the book. Tal was saying that only ONE move would have held for Smyslov PRIOR to >the sacrifice. Genius's evaluation stayed for 2-3-5-10 moves... Then suddenly, >it said Smyslov was losing. Genius would have played all the same moves Smyslov >did, after the sacrifice. Now, I let Genius calculates about 3 hours on the >position PRIOR to the sacrifice and it was still never seeing it coming, since >it thought it was bad for Tal! So, maybe avoiding tactics against a computer if >you are a GM known to be good at tactics is not a good idea? By the way, if >Kasparov had played very tactically against Deep Blue (playing as he did in all >his carreer), maybe he would have had blasted it too? Cause if Kasparov has the >advantage in a certain position, he can't lose it without a mistake. Now, what >percentage of his games does he blunder? Normally, the higher the rating, the >less blunders you see... exaclty.. the higher the rating the lesser the blunders.. so why is it that aa 2500 rated computer program makes positional mistakes? If it could be brought not to make positional mistakes while retaining its speed at tactical eveluation (with a dual processor system) it would do much better. > >Serge Desmarais
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.