Author: Reynolds Takata
Date: 21:58:03 01/25/99
Go up one level in this thread
On January 25, 1999 at 21:26:28, James Robertson wrote: >On January 25, 1999 at 19:19:22, Reynolds Takata wrote: > >>On January 25, 1999 at 18:36:40, Howard Exner wrote: >> >>>On January 25, 1999 at 15:59:07, James Robertson wrote: >>> >>>>On January 25, 1999 at 14:04:39, Reynolds Takata wrote: >>>> >>>>> >>>>> >>>>>A few months ago, a poster named Gram or Graham can't remember, said that Fritz >>>>>5.32 playing anonymously could score the GM norm, well it turns out he was >>>>>right. >>>>> >>>> >>>>Not necessarily. Humans play a lot of moves against other humans they would >>>>*never* play against a computer. >>> >>>I remember Shawn's thread and that was his point. That if the computer >>>had some disguise (human cheater) then humans would play it as if >>>they were playing a human. All the anti-computer play would not occur, >>>as you have just said. So the computer playing anonymously would give >>>the machine an advantage in the sense of stripping computer savy opponents >>>of their arsenal of tricks. >> >> >>I just got an email from Shaun about the post, he thanked me for posting it, but >>said that his real point was that Computers are Grandmaster "strength" against >>regular "human play", though against anti-computer play not necessarily so. He >>went on to say was that anti-computer play was a "sort" of cheating > >Cheating...? ? ???? I can think of no human GM who does not adjust his play >depending on his opponent. To say that doing the same thing against a computer >is cheating is absolutely ludicrous. > May god bless you my son, and all of the happiness the world knows shine upon you. If you would care to make note to the above, you might note that the word sort is in parentheses, this is an obvious and clear indication that cheating in the strictest sense of the word was not meant. What was being said is that because humans can get "definite information" on a computers play, as opposed to "likely" information about a humans play, that this "studying up" on the opponent in this case skews the result more than when one studys upon a human. This for the reason, that a human opponent isn't necessarily going to play like their record might indicate that they are going to play, and further that most often the human has the option, to study the computer. "Case in point. Here in the great state of Hawaii i have a friend who has Rebel, i saw a game were rebel was defeated against by a strong GM. I played the program at my friends home, and won the exact same game! He thought i was Kasparov, his computer never before having played the line, didn't have it stored in it's learning file to avoid the line. Though i'm a master, i could have just as easily been a 1200 rated player, and done the same thing especially because the game opened 1.g4, and i expect that any new copy of Rebel would play the same line. The same thing could happen in a tournament. The computer as you say not being a living entity itself can does not have the ability to go out and search out the specific humans oponents weaknes. Thus it can quite fairly be argued that this poses a program a disadvantage, not an illegal disadvantage, but a disadvantage, thus as i said "sort" of cheating is simply a description of this disadvantage. > >>against >>computers. He gave an example of giving a weaker player(competent and strong >>still though) a detailed description of all of Deep Blues weakneses that might >>be garnered from a log of a 1000 DB vs DB games. Deep blue would have no >optionto change its nature or change as a human would. > >This is wrong. The program is not an "entity"; rather it is entirely it's >programmer's work. The programmer can change the program however he wants, as >much as he wants, whenever he wants. For pete's sake, he can even upgrade the >hardware, an option not available to the human! Deep Blue is not a living entity the whole point of the above, you obviously missed it, but we all miss things from time to time, i'm sure this was just a lapse. If i knew Rebel 10(the one out of the box) was going to play in a Swiss tournament, and i got to play it, i could have well have prepared for it, and Rebel would in no way be prepared for me, the "programmer changing the program isn't always an option". The point was that the program "because it's not an entity" would just play unawares of my strengths. I play and beat my programs all the time, because i know the weaknesses, this is no definite reflection on their strength. Though the book learning features of some are keeping them from falling over and over into the dead bust line. Though an 11 year old kid beat rebel by copying a game i won because of a transposition that occured late down the opening. Kid is rated 1275, I take it that sounds like no disadvantage to you huh? > >Also, I remember a huge number of outraged posts that Fritz was learning against >specific opponents and winning more games later in it's SSDF matches. > >>More to the point he said >>"imagine that the human cheater(disguised comp), not necessarily ina tournament >>sat down before 10 grandmasters for a 40/2 on ten different days(unbeknownst to >>the GM's), the comp might defeat all of the GM's or the majority. So how could >>one say that the comp isn't GM strength". > >Back to my original post; preparing for your opponent is paramount to success, >even against humans!! It is? "PREPARING FOR YOUR OPPONENT IS PARAMOUNT TO SUCCESS"(James Robertson 1/99 CCC). Now even i didn't go that far, but i did just say that because computers generally don't prepare for opponents yet their opponets prepare for them, that computers are often at a disadvantage. SO.. what are you arguing about??? >>What can be said is that in a tourney a computer is often at a disadvantage, >because his opponent knows the computer, but the computer doesn't know >anything about the opponent(a disparity). >>Both Chessbase(nixdorf classic), and Rebel(anand match) understood this, and >>attempted to make their programs play openings that their opponents had >>previously had difficulty with. The result of giving the comp knowledge of the >>specific opponent gave Fritz a HUGE success. I believe this is also one of the >>main reasons for Anands defeat in the blitz match. If not his defeat, at least >>his getting BLOWN OUT. I'm certain giving comps knowledge about how to avoid >>certain types of positions(which is already being done), will incerase comp vs >>human scores considerably(even though by my count they seem to be practically >>winning now anyway). > >This doesn't match my count..... Well it matches mine...... > Have a nice day, and may doves and rainbows be on your path. R. Takata USCF Life Master Fide Master > >>All this is by the comp is an attempt to "Accentuate the >>possitives in their game" and cover up the weaknesses by playing for what they >>like. Humans do this everyday in chess. No one who hates open positions, >being >>down material and attacking is going to play the Ruy Marshall. Hey i'm tired >of >>writing :). >> >>R. Takata >>USCF Life Master >>> >>>>Imagine Tal sacrificing in a tournament filled >>>>with computers; an ugly sight. >>>> >>>>Also, we have no clue what the hardware was. >>> >>>Yes, that would be usefull info in knowing the hardware speed. >>>> >>>>James >>>> >>>>> >>>>>>The German Newsmagazine "Der Spiegel" today reports a funny story: Mr. >>>>>>Allwermann, an Elo 1925 amateur of age 55 has won a nine-round 2h/40 swiss >>>>>>tournament and achieved a performance of 2630. >>>>>> >>>>>>Organizers and competitors got somewhat suspicious when the guy announced a mate >>>>>>in eight in the decisive final round game against grandmaster Kalinichev! >>>>>> >>>>>>"Der Spiegel" writes that Mr. Allwermann's moves are reproducible with >>>>>>Fritz5.32. While nobody understands how he has done it, there are rumours that >>>>>>he formerly worked in the 'electronics business'. Moreover the German chess >>>>>>magazine "Schachmagazin 64" not only points out the fantastic attacking >>>>>>combinations but also some typical Fritz 'no-clue' moves like Bf4 in a closed >>>>>>French Winawer as White. >>>>>> >>>>>>Seems like we will need airport-type security checks in tournaments in the >>>>>>future. >>>>>> >>>>>>Jürgen
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.