Author: blass uri
Date: 04:37:06 08/09/98
Go up one level in this thread
On August 09, 1998 at 06:57:49, Robert Hyatt wrote: >On August 08, 1998 at 19:16:57, Mark Young wrote: > >>On August 08, 1998 at 18:01:58, Robert Hyatt wrote: >> >>>On August 08, 1998 at 17:00:38, Leon Stancliff wrote: >>> >>>>Bob, >>>> >>>>I posted a reply to your previous message concerning standard GM vs computer >>>>games. But since it was buried back a day or so ago, I decided to post a new >>>>message. >>>> >>>>I would appreciate it very much if you can provide us with as many as a half >>>>dozen or more of Kim Commons games vs computers at standard times. >>>> >>>>I will also add that your contribution to the dialogue on ICC is both >>>>interesting and stimulating. >>>> >>>>Leon Stancliff >>> >>> >>>here is the first "group" of 20 10 games Kim Commons played against Crafty on >>>ICC. I simply took the games in order of play. There are more if you want to >>>see them... >>> >>>Bob >>> >>Crafty Scored 7wins 2losses and 1 draw in the last 10 games played by IM >>Commons.(i guess the last 10) This is a player who knows how to beat chess >>computers, and I'm sure knows crafty as well as any player. And the games were >>play at fairly slow time control for ICC. I find it hard to understand why you >>think crafty is no better then 2400 at 40/2. I just can't see IM Commons playing >>that much better at 40/2 and crafty just falling apart. I think crafty would >>still win play IM Commons at 40/2. >> >> >> > > >I understand your "question". My answer is simply based on what I know about >Crafty and what it "knows" about chess. Kim is quite strong. Most of the >time, he plays "chess", the games are interesting, and anything can happen. >On occasion, he plays "anti-computer" and it is *much* harder to beat him >when he does. I'd suspect that in a real game, in a real tournament, I would >not see some of these openings that castle opposite and both sides push pawns >and so forth. > >But the real reason I stick with my opinion is that I can see all the things >that an IM or GM knows about the game that Crafty doesn't. If I try to add >all the special case stuff that the Deep Blue guys did, I'd run at 100 nodes >per second, *if* I could figure out how to add the knowledge in the first >place (assuming I could even quantify this "knowledge" of course.) > >IE in the second Rebel Anand game, I showed that position to a GM. He >looked at it about 30 seconds and said this is easily either winning or >drawing. I asked "do you really think those pawns can't be stopped?" (or >something to that effect). There was a long pause, then a "no bob, you >musunderstood... black is almost certainly winning, or at least is >drawing, unless I have overlooked some forcing tactic that I don't see." > >So the GM "knew" that black was playing for a win, and could at least >expect to draw, while every computer I have seen says white is better and >would play right into that line. Does that sound like a GM program? Or >does it sound like a program that is trying to use impressive tactical >skills and a mediocre evaluation to cover up the weaknesses it really has? I am not convinced by this becuase GM programs do not have to understand what GM's understand. There are many positions 1800 players are better than fritz5 and I believe even better than deeper blue. For example computer programs do not understand the idea of fortress positions when one side has big adcantage but s(he) cannot do nothing with it. The only question is if the programs can win GM's over the board Junior did 3 draws against GM's in the last serious games it played against GM's so I believe they are not far from being GM programs. I believe 2500 is the level of the best programs. Uri
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.