Author: Jeremiah Penery
Date: 23:37:44 11/30/99
Go up one level in this thread
On December 01, 1999 at 02:16:11, Tom Amburn wrote: >On November 30, 1999 at 23:33:21, Jeremiah Penery wrote: > >>On November 30, 1999 at 21:34:54, Charles Unruh wrote: >> >>> In the last 3 days i have seen at leaset 6 posts trying to make out that the >>>progs were barely USCF master strength! There is no mere master in the united >>>states that could dream of beating sherbakov with money on the line, or beat the >>>lithuanian national team, or beat Gelfand in a 40/2 it does not happen! Yeah >>>people are entitled to their opinions but i think there are limits come on. >> >>One game doesn't really prove anything. >> > >Sorry but i do not think you read charles's original statement. He gives more >than "One" Game, but atleast twenty! Gelfand played two games against junior as >well as the over a dozen games of rebel against I'ms and GM's. I was specifically referring to Junior vs. Gelfand, which I thought was one game. So far, what is the score of humans/programs in these games? >>Take the following two scenarios: A USCF master vs. Gelfand in a 40/2 match of >>arbitrary length; Rebel (or another program) vs. Gelfand in the same type of >>match. This is what will likely happen. >> >>Match 1 -> Master vs. Gelfand. Gelfand will probably win most of the games. Of >>course there will be some draws, and the master will win a few, but this ratio >>will stay fairly constant through the match. > > > This is pure conjecture! ? What do you think would happen in such a match? Notice that I said nothing about any program here; I'm talking about a Master (USCF, FIDE, whatever) human player vs. a GM. > On the other hand We have the factual results of Rebel >vs Grandmasters which proves Rebel is Much stronger than 2200, which I think is >what Charles post is all about. Again, one game, or even 20, is a far cry from "proving" anything. >>Match 2 -> Program vs. Gelfand. Gelfand will initially probably lose some. At >>this time control, the GM has time to either calculate the necessary tactics, so >>that he doesn't get crushed that way, or he can simply keep tactics out of the >>game, for the most part. The program will still win some, however - maybe just >>as much as Gelfand wins. After a while, Gelfand will find flaws, holes, in the >>program. He will then exploit these. Gelfand will start winning more. >>Eventually, he is winning as much (or more) as he did against the master. >>Take, for example, the first DB-Kasparov match (1996 - NOT DT-Kasparov.). By >>the final game, Kasparov had already done this assessment of the computer's >>strengths/weaknesses. It could be clearly seen from the game that Kasparov had >>adapted to take advantage of this, as any GM will eventually do when playing a >>computer. >>Since, IMO, DB is a fair bit stronger than any micro program today, this will >>happen even more easily to the micros in a 40/2 match against a GM. >> >>Have you ever watched the computers play on FICS/ICC? Watch some games between >>lower-rated players and them. Would a GM, even an FM (FIDE Master), _ever_ even >>draw to a 1400 player? No. But the programs do all the time. Even I, a fairly > > >The fallacy here is that you never know if the 1400 is using a computer or not. What are you talking about?? I'm talking about OTB play, as in a tournament match of some sort. >I am over 2100 and lose often to programs although I do win occasionally. I tell >you this I know for a fact no 1400 player can draw a computer. Oh? Did you see the game "Garf" vs. Chess Tiger from FICS that was posted? In that game, Garf was rated 1100 or something, but was able to draw Chess Tiger, which is supposed to be the strongest computer program. This is just one game; there are tons of other examples. As I said, even I can occasionally win/draw against my computer running whatever program (Crafty, Rebel, Hiarcs, etc.). I am probably a <=1400 player. >>bad chess player, can get the occasional win/draw against the computer. Against >>a Master human, I would _never_ have such chance. >> >>The conclusion is that after a sufficient amount of games, the GM can learn how >>to play against the computer. He can figure out the exact strengths/weaknesses, >>and then play them accurately to his advantage. He can't really do this against >>the human. >> >>So while we have relatively few games of programs vs. GMs at 40/2, the programs >>will appear very strong: they will win as much or more than they lose. If the>>GMs begin playing computers regularly at 40/2, in regular match conditions, I >>think the GMs will come out clearly on top. > >You are probally right , I doubt computers are yet GM's but they are real close. I think it only looks this way so far. I think if the GMs play computers regularly enough at 40/2, they will eventually learn how to play against them, and start winning a lot more than they are now. Thus the computers' ratings will go down with respect to the GMs. >But still I think charles was disputing the allegations that computers are not >even 2200 players, which is borderline lunacy. I don't think it's lunacy. As I said before, a 1400 will probably never win/draw against a GM, but they do all the time against computers, even in blitz games. In 40/2 games, it's even easier for humans.
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.