Author: KarinsDad
Date: 14:49:28 02/03/99
Go up one level in this thread
On February 03, 1999 at 16:43:29, Matt Frank wrote: >On February 03, 1999 at 13:13:58, KarinsDad wrote: > > >>Matt, >> >>I have read the Hiarcs7 vs. GM posts with interest and it suddenly dawned on me >>that the majority of chess players in the world DO NOT analyze their opponents >>games ahead of time. GMs and IMs do it and some local players may do it against >>each other, but that is it for the most part. > >For matches between humans it is almost invariant at the GM level. Only for certain types of matches such as candidate or world championship matches. For example, there were this many GMs in the following events: 26 GMs in 59th Hoogovens 8 GMs in Lost Boys 5th 12 GMs in Fontys 97 6 GMs in Hoogovens USA 26 GMs in 60th Hoogovens 14 GMs in US Championships 98 22 GMs in 61st Hoogovens Are you telling me that these GMs have the time to prepare against every other GM in one of these tournaments, for white and for black? A lot of times, substitute GMs will be invited at the last minute. Yes, GMs study the games of other GMs, however, even they do not have an overwhelming knowledge of the style of play for more than a handful (15 to 20) of the GMs that they play often. Anything else would be overkill. Stating that a GM should prepare against a computer because GMs sometimes prepare for some of their human opponents is not realistic. The true reason that this "excuse" is used is that GMs are treading on thin ice with computers (they do not know exactly how much better they are getting each year) and so they want an assurance of a possibility of fairness (i.e. the chance to study and practice with the program). It's a security blanket type of thing. >> >>To be totally FAIR, you should have never told the GM which computer program he >>was playing (you could have even mislead him by saying a new program). If you >>would have done that, the GM could not have prepared for Hiarcs 7 specifically >>(just computers in general) and you could have not prepared with Hiarcs 7 for >>the GM (i.e. just people in general). > >To be totaly fair to the GM he needs to know if it is possible to prepare >himself as against any opponent in a match. And as Dann has said on a previous >post these matches are more likely to occur if GMs feel that they won't get >sideswiped. Granted, a GM may feel more comfortable preparing ahead of time. My point is that the GM can practice against a computer whereas a computer cannot practice against the GM, therefore, allowing both to prepare gives the GM an advantage. >> >>This analyzing of style to gain an advantage is kind of bogus. It cannot be >>helped at the GM level for human vs. human play, especially in this day and age >>of databases and NICs, but to have an unbiased test of the computer vs. the >>human, it would have been better to use a double blind. You would have tested >>two non-poluted entities. > >That is a pure test of the system that I think is very hard to control for >(because of concerns by the human particpants It would not have been if you would have kept it a double blind to begin with. If during your negotiations with the GM you said, I will not tell you which program, but on the other hand, the program will not prepare either, would have been fair. If the GM balked on this, so be it. You could always re-negotiate with the arrangement you currently have. >> >>You are not testing whether Hiarcs 7 plays at GM strength, you are testing >>whether a tweaked version (i.e. changing the opening book) of Hiarcs 7 can play >>at GM strength against a GM who is trying to find flaws within Hiarcs 7 ahead of >>time. > >So when humans adjust for an upcoming match what would you consider that >> >>Instead of two warriors battling, it's like two librarians battling. > >That is not the way it will look in June. Trust me. Librairians don't get bloody >:-). If the GM crushes the program, it will be a battle of researchers (and hence the reason Bobby Fischer and other GMs would like to see random chess). >> >>For example, if the GM finds a series of moves that leads to a lost game (or a >>series of lost games based on different choices in the opening) in Hiarcs 7 and >>manages to get Hiarcs 7 to play it, what has this proved? That the chess program >>is partially deterministic? > >Hiaracs 7 has lost some games that when analyzed will show they were draws or >possibly wines in some situations. It is just such preparation prior to a match >that makes for new and exciting chess. >> >>As I said, it is kind of a bogus test (interesting and anticipated, but bogus). > >I will agree that the purest test would be Hiarcs 7 versus GM sight unseen by >each side. That would have been the optimum situation for the computer. However >it is not meant to be. Something to think about for future matches. KarinsDad :) >> >>KarinsDad :) > >Best regards, >Matt
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.