Author: Robert Hyatt
Date: 19:24:54 07/05/01
Go up one level in this thread
On July 05, 2001 at 17:31:41, odell hall wrote: >HI CCC > > On many occasions at this forum, i hear some say that the human lost because he >or she, failed to employ some undefined Anti-Computer Chess Strategy. I think >however such statements are very presumptous, I suppose that many of the people >that make these claims cannot defeat the computers themselves, using any >strategies, including anti-computer. If the human loses, then the excuse is >always lack of anti- computer chess play. The problem with this line of thinking >is that many of us ourselves don't really understand what anit-computer >strategies are, ourselves, and would not recognize these strategies even if it >were employed. For instance, Dr Hyatt is one of the number 1 people here who >allegedly "understanding Anti-computer play" yet he admittedly never beats his >own program???? IF such strategies actually reduced the strength of computers >then wouldn't Dr, Hyaat be able to regulary defeat his program?? He cites Roman >as an example of successful anti computer play, yet roman wins less then 50% of >his games versus crafty. Most everybody believes that "blitz is solved". Yet Roman can win near 50% at times, and even over 50% on occasion. Against _any_ program on ICC, not just against Crafty. And this is at 5 3, the program's supposedly most difficult level to win against (blitz). How do you think he might do at a longer time control? If I play serious chess, I can probably come up with 1 point every 20 or so games against Crafty. If I played regularly for a couple of years, I could probably bring this to 1 point in 6-8 games. I can outplay most programs positionally, but as the saying goes "meat makes mistakes". I have beaten gnuchess many times on the servers, playing anti-computer (closed) lines that lead to long-term kingside attacks. My main problem is that I simply no longer play much chess at all. Rust hurts... As far as "is anti-computer chess always doable?" In general, yes. A program can always give up a quick pawn to try to keep things open, but most such real gambits (ie the Evans, the Latvian, and so forth) can lead to bad positions if the gambiteer isn't careful. And even then, white has to ask black to play the latvian by playing 1. e4 of course. The problem is that at the present time, due to FIDE circumstances, GMs have little reason to really study computers and prepare against them. Even if they play in a tournament with Roman and they get rolled by a comp, and he beats it easily, there is little incentive for them to try to rectify this since they know the event is unrated, and that the computer can _never_ win a prize. If that ever changes (and it probably never will) then the GMs will bounce back. A few might one take "take up the gauntlet" and try to learn how to handle the silicon beasts... But most probably won't since there is no "payoff" for investing the time... > Personally i believe there are anti-computer >strategies, but they work only in limited positions, And openings, a human >cannot guarantee that the game will be steered always in the anti-computer >position they desire. Which is why Anti-computer strategies as a whole have >failed on the tournament level, computers are consistently scoring even better >then their SSDF Rating suggest. Even those who are sopposed to really understand >anti-computer Chess are losing, like van der weil, and also Roman. The most >prominenet example being kasparov even after months of "TRAINING".
This page took 0.01 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.