Author: KarinsDad
Date: 15:38:36 02/03/99
Go up one level in this thread
On February 03, 1999 at 17:39:35, Matt Frank wrote: > >>1) The GM can practice against the computer. The computer cannot practice >>against the GM. > >By playing against the GMs' games the computer is in effect playing the GM (in >preparation of a match or game). Not true. If I was a GM in your match, I would stay away from lines that I have recently played (last 10 years or so) and go back to lines I played in the past or lines from games that were not published. Most variations are fine as long as you are familiar with them. I would refamiliarize myself with lines that the computer (i.e. it's learning capability) would not expect. I could then play those lines against the computer and see how I do (i.e. to see if I know them well enough to win). Note: these would have to be solid lines and not some of the stuff that Kasparov was playing. For example, as white, I could play 1. Nf3 to prevent e5 as my opponents first move. My intent may be to play 2. c4, but to avoid variations of 1. c4 e5 with a reverse Sicilian. At the GM level, he can fairly well control which openings can be avoided. The computer has a more difficult time of doing this since it would have to have a GM or IM and a programmer go over it's opening book to ensure that it avoided lines that it's programming did not play as well (such as the conversations here at CCC on the French and the King's Indian). > >>2) The GM can modify his play during a game and during the match. The computer >>cannot. If the GM finds a series of lines where the computer repeatedly loses in >>practice, there is a good chance that the lines will be played during the match >>and the computer may be doomed to lose at least some games before the match even >>begins. > >Some software is designed to modify its play based on its opponent. Yup. This is a chancey thing for the GM at best. > > >>3) The GM has the example of where Garry failed against Deep Blue to show him >>how to NOT play against a computer program. Do not try to lead it into inferior >>positions. Always try to play the best move on the board and not necessarily >>anti-computer moves in all cases. Garry probably learned more about how to play >>against a computer from that loss than even he realizes. This GM has more of an >>advantage than Garry had against Deep Blue since this GM can analyze previous >>Hiarcs 7 games and can also practice against it. Garry did not have these >>advantages. He had to guess based on other computer programs. > >Thats why a Hiarcs 7 vicory would be so impressive. True enough. > >>These differences are significant and explain why the experiment is skewed. The >>only FAIR test would have to keep both parties in the dark. And that could still >>happen if the GM agrees not to analyze Hiarcs at all and Matt agrees to use the >>standard opening book. But they are probably too caught up in their previous >>agreements to make that adjustment. > >Further, related comments are made by Ed Schroder the programmer of Rebel where >he says that it is easier to prepare against Humans than computers. Odd >statement, you might say, KarinsDad, but based on his 18 years of tournaments. Wonder why he said that. Any ideas? Probably since humans are more deterministic. KarinsDad > >Matt Frank > >>KarinsDad >> >>> >>>James >>> >>>[snip]
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.