Author: Serge Desmarais
Date: 16:50:30 09/04/98
Go up one level in this thread
On September 04, 1998 at 13:40:56, Fernando Villegas wrote: > > >Hi all: >Many times it has been said here, specially by Bob Hyatt, that there is not >transitivity in terms of strength between programs: If A beat B and B beat C, >that does not means, then, that A surely beats C. This is true and the same is >true, as experience show, in the realms of human players of any category. But >then I wonder if also is true in the mixed field of programs vs human Vs GM. I >wonder if the known fact that GM are still stronger than programs in long games >and the also known fact that top programs are far stronger than A and even >expert level players, then that necessarily means that GM are stronger than top >programs in relation with those low level players. >My feeling is that this is a debatable issue. In fact, I even have some >arguments to say that maybe top programs are stronger than GM in that specific >relationship with low level players. Let me explain my point. >What makes a GM a GM and not just a master or IM is specifically his superior >positional grasping of the game. They see, in that field, things that we does >not even suspect. In the most rarified layers of that stratosphere, say, in >games between Karpov and Kasparov, we just don’t understand what is happening, >why one of them resigned or offered a draw; they see things far beyond our chess >understanding. >Nevertheless, it is very unlike that a game between a GM and an expert o A class >player could ever reach a depth where such kind of subtile things becomes >necessary. The weaker player will be finished long before just on the ground of >tactical shots or, more likely if we are talking of an expert player, with >positional pressure of the kind you can meet in a Master or IM level player. >That is more than enough 99% of cases. So, with this low level players, the >knowledge and expertise of GM in the most subtile areas of positional >understanding will stay without use, or, in other terms, will become marginally >useful for the strongest player. >But then, look what happens from the point of view of the weaker player. When he >is playing -in a simul or, like in my case, in my home with a friend GM- he >surely will be defeated, but as much the GM is nevertheless a human being, there >is a likelihood that maybe, maybe, MAYBE you will be capable sometime to kill >him with a tactical bullet in the head before he can use his superior positional >understanding OR, more probably, that the GM will miss a tactical shot you made >possible due to an inexact move or even an straight blunder and do you will be >let playing with a hope to get something. And you know it. You know you can do >it, that you have the chance. You know that maybe a minute imprecision will be >not detected after all. >But against top programs the feeling is entirely different. What you know from >the beginning is that nothing of all those tactical tricks you learned after an >entire life of chess playing will be useful at all; you know you are not going >to kill the program with a sudden pin, a mate threat in the last file, a double >attack or a 5 to 6 ply pretty combination. Even GM sometimes fall to those >traps, but never a top program. And besides you know thought bitter experience >that the most insignificant imprecision will be duly and severely punished. In >fact, the normal experience we have playing a top program -IF you are not more >that a expert player- is this: you are holding the game, you are playing pretty >and precise moves one after another, you see lot of threats and keep them at >bay, you even develop certain threats against the program, you get a nice >position and then, sooner or later, you commit even the most microscopic mistake >and immediately the program punish you with a blow and your nice, promising >position, fall like the house of Usher described by Edgar Allan Poe. >It is not so? It is not truth that after many experiences of that kind the >intermediate player I am talking about tends to feel impotent and extremely weak >and, in fact, almost without chance against top programs? And is not that >feeling equal to a awesome feeling to be facing a far beyond stronger player, >far even that a GM but human player after all? >Of course all this is sheer speculation as far many players does not have any >experience of playing GM’s. No way to make a measure of this and prove with >numbers that top programs are, in comparison with weak players, stronger that GM >against he same opposition. All this is just an speculative statement about this >issue, maybe a way to add a new element to the complex field of relative >measurements of strength and ratings. >Fernando It is an interesting idea. Though, I do not share it. First of all, how would you measure that a chess program/computer is better than a GM against amateur players (A-class or experts). By the average number of moves in the games? From my personnal experience, it is just the opposite that I have seen! An expert player of my friends defeats one of the SSDF top program (Genius 3.0) about 1 time for every 3 games, while a Master I know wins about 2 times in every 3 games against the same program. I did make several draws against the same Genius 3.0 program myself (an A or B-player) but NEVER against these 2 players, though I did not play a great deal of serious games against them. Also, it is not just their superior positionnal knowledge that makes the GMs what they are. It is also their superior opening knowledge, their superior tactical calculation's habilities, their superior ending knowledge etc. In tactical positions, they calculate more precisely and deeper than the weaker players (take a look at Kasparov's games, as well as those of Fischer, Anderssen, Murphy, Tal, Janowsky, Marshall, Spielmann, Alekhin). Computers are known for their exceptionnal tactical habilities : if a forced mate exists in their horizon, they never (almost) miss it! Same thing about if they can win/trap your queen or any other piece. Moves, to which several annotators have added one or several "!", just take a few seconds for modern programs to "discover" and prefer. But at long tactics or "true" sacrifices, computers aren't too good. In one game Tal sacrificed a piece against Smyslov. In the book (was one of a friend, so I don't have it here) Tal was giving Smyslov as being lost after the sacrifice. I have had Genius look at the positions from about game moves after the sacrifice and for several hours. Almost all the way, the program was seeing SMYSLOV as winning, until after about 10-12 moves, it SAW that Smyslov was lost! Then, I went back, letting the program "think" for about 1 hour on the position PRECEEDING the sacrifice. It was preferring the same losing move as Smyslov did! Here is another example : Bronstein playing a chess program (would have to search to fing which one) in an Aegon's tournament. Bronstein sacrificed a piece and blasted the program tatically! I tried the same position on my programs and, as I remember, they saw nothing good for Bronstein in that sacrifice, though later they admitted he was winning. What is fun in that is that Bronstein just took a few seconds (at most 1 minute) to make the winning sacrifice. After the game, he just said he KNEW it was winning and did not have to calculate anything! In fact, he had so much experience with such positions that his brain/instinct was telling him that it was winning! So, some GMs are so good at tactics that they don't need to always calculate everything when in familiar waters. I myself had a great win against Chessmaster 4000 in the first half of the 90's. It claculated a very pretty combination that took me by surprise and won one of the pawns on my castle : I hadn't even seen that! After that, tripling the pieces on the g-file and, I got the initiative and the advantage and managed to win! There was also a nice sacrifice I made against a weak computer Advanced Chess Challenger Voice), in 1984, that gave me the win. Accepting the sacrifice means losing : it takes 12 plies for Fritz 4.01 to not accept the sacrifice (though even for me, who is not even an expert, it is EVIDENT just at the first sight that it is not good at all to accept the sacrifice! The only way I see that a GM could have trouble against an amateur player is if he underestimate him, not paying too much attention to the game nor taking his opponent seriously. Serge Desmarais
This page took 0.01 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.