Author: Robert Hyatt
Date: 15:19:54 07/13/98
Go up one level in this thread
On July 13, 1998 at 14:52:02, Shaun Graham wrote: > >>> >>> >>>No Hyatt your logic is flawed. First of all i didn't say anything about as you >>>put it "MANY GM's", The truth is however that there are some and i would gather >>>a good number. Another thing when people use that statistic about how many >>>times you should statistically beat someone, it really isn't considering that if >>> I am playing a certain number of games against a single person, or wether i'm >>>playing against several people of that rating and what the effect is. I myself >>>have just recently played a match with an INDIVIDUAL 200 points below me , i >>>beat him perfectly, , besides that, i have beat him 14 times straight! I would >>>bet that's a bit beyond statistical error. >> >>your point would be? if you beat him less then 3 of every 4, I'd be concerned >>that his rating was too low or yours was too high. As it is, either yours is >>too low, or his is too high, because if he is winning none, he is way over 400 >>Elo points worse than you... > >No he is not 400 points below me in fact he has been improving very fast. i >beat him because i'm better. plus read what I wrote again. *IF* you win every game, you are *at least* 400 rating points better than him. *period*. That is what Elo is all about. >> >> >>> >>>Further i never said that computers weren't getting busted badly, because they >>>are, but you simply overlook anti computer strategy. As i said if i played in a >>>Swiss system tourney recieving moves from fritz, without people adopting their >>>anti Computer strategy, they would play normally, allow the positions to get >>>open,play for tactics, and then bye-bye i'd have the norm. Further i wasn't >>>talking about computers in general, i was talking about Fritz. I already know >>>or i shall say that i've heard, that you don't like the way fritz plays. >>>However to pull your own tactic, according to the STATISTICS of both selctive >>>search, and SSDF fritz is the strongest program. Further people who understand >>>chess much better than you, me and most people GM Yermolinsky, and i also >>>believe Anand believe Fritz to be the strongest commercially available program. >>> >> >> >>it probably is the strongest available. But it is also the worst at defending >>against "anti-computer" strategy. And whether it would do better "incognito" >>isn't an issue, because it can't play like that. > > > The point was never to make it an issue, but to show that you can't truly know >what type of result a Chess program would know without computer bias by the >opponent. > but who cares? Because such an event is completely impossible anyway. YOu can *not* play GM players in a FIDE/USCF/etc event "anonymously" where they won't know you are using a computer. That is totally impossible. I'd bet that a computer could tear hell out of a GM if they played in the middle of a rock concert too, because who could concentrate? The computer sure could. But it won't ever happen... >And if it can't, >??? Thus by this admission are you admitting that without anti computer chess >techniques it would win? nope. Kasparov didn't play anti-computer in the first match and still rolled Deep Blue pretty well. But given the choice, most players would play the style that gives them the best chance. And against a computer that is a closed position. > > it is going >>to get eaten alive by GM's... > >What > >> >> >>>Of course neither you and i can prove wether fritz is a GM strength or not, >>>without actually testing fritz in actual tournament play(I believe that this >>>would also have to be done just as i said in a secret way with an individual >>>recieving moves from fritz, to avoid bias, so that players would play the way >>>they do normally). It is none the less, my belief that indeed Fritz(current >>>version) could perform well enough, to eventually recieve 3 GM norms in 5 >>>Years(the time in which one must obtain all 3 norms)playing on the europeon >>>swiss circuit of chess. >> >> >> >>That's an invalid way of testing. > Everything has to be out in the open, > >Quite incorrect!! It always depends on what you are testing, and what you are >trying to find out. In this case since it is known that knowledge by the >opponent that he/she is playing a computer effects how they normally play, you >would never know how good at playing against standard chess a program is. What >you WOULD BE TESTING, is how the program plays against anti-conventional methods >of play(anti-computer strategy). we are talking about a computer "GM" are we not? In that case, there is but *one* way to test... in tournaments. That's the *only* way to get a GM title. And it makes your argument "moot". > >>human GM has to know he's playing a computer (although if you make the match >>last 24 games, Fritz would get destroyed no matter what because the GM would >>"figure out it's a computer" after just a very few games, and then it would be >>over... > >Hyatt you may be a computer scientist, but apparently your education in >experimental scientific technique is limited. If you are trying to test a >programs to find its strength in real tournament chess, against normal chess >play. Then it is of UTMOST IMPORTANCE to eliminate the bias. When a player >knows he is playing a computer, the tester of such an event can not know that >this information is not skewing the behavior of the player, and thus the >results. So you can not know how strong the program is against standard chess >play. What you would be fnding is that against anti-typical chess play, the >program does not perform effectively. Further, If a person was playing an >opponent with no expectation that they were a computer, NO, they would not think >it was a computer. If i was feeding Shirov moves from Fritz, for his match >against Kasparov. Kasparov would never think shirov was cheating! He would >just think that shirov was playing considerably worse chess than usual(for the >reason that kasp and shirov are of course much better than fritz or the average >GM). Just for arguement sakes, say the grandmaster did find his opponent was >using a computer, again you would be observing how a program plays against >anti-computer chess, something that does not occur in standard chess play. My scientific method is a lot sounder than yours. Because you are trying to create a whole new class of GM player, the "anonymous GM". It won't happen.
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.