Author: Robert Hyatt
Date: 13:13:41 09/08/00
Go up one level in this thread
On September 08, 2000 at 13:30:50, Walter Koroljow wrote: >On September 08, 2000 at 00:28:13, Robert Hyatt wrote: > >>On September 07, 2000 at 12:34:40, Walter Koroljow wrote: >> >>>Thank you for the link. It turns out that Zugzwang ran on a multi-processor >>>Cray! Therefore, I have excluded it from consideration. >>> >>>That leaves 29 program/PC combinations that played 157 games (+73,=57,-27) for a >>>score of 101.5-55.5 (64.6%) against opposition with an average rating of 2415. >>>A single program with a rating of 2541 would be expected to achieve this score. >>>Recalculating the confidence statistics, we get only minor changes: >>> >>> 95% Confidence Interval >>>Spread of Ratings for Average Rating >>>----------------- ----------------------- >>> 0 2502-2579 >>> 200 2502-2584 >>> 400 2503-2595. >>> >>>The conclusion stays the same: There is considerable confidence that the >>>average rating of the programs in Chris' data running on 200MHz or faster >>>PC-based computers is in the 2500's. >>> >>>Cheers, >>> >>>Walter >> >> >>You might conclude that if you look only at the _results_. But if you look >>at the _games_, you get a significantly different "picture". A GM simply won't >>get beaten by a 1800 player. A computer falls into such losses. A GM doesn't >>play like a genius in 2 games and then like a 1600 in 1. > >The picture if one looks at the games seems very complicated to me. Some recent >examples: DJ6's horrid ...Ne4xNd2?? against Kramnik; FritzSSS crushing Van der >Doel (2537) in 15 moves; and De Vreugt (2498) demonstrating complete ignorance >(9.c4??) of how to play against computers. I can't begin to generalize from >these and the other few anecdotes I am aware of. But you have had many years of >observing human-computer play. Have you observed systematic patterns or trends? Several. Computers are still overly materialistic. They will often win a pawn and totally wreck their position. Or win a pawn heading into a position that a human knows can't be lost by the other side. And endgames are _still_ not solved. I can give you a couple of cases I watched another Computer play against Crafty last night, while an IM and I were chatting about the games. Computers are still horrible at overlooking kingside attacks. And they often don't understand blocked positions at all. Trying to avoid blocking things is one 'crutch' that I rely on with Crafty, but it isn't the "final solution" by any stretch of the imagination. > >And what lies in the future? The statistics indicate that computers have done >well. Some things could change this picture. On one extreme (low effort), the >De Vreugt example shows that humans could do better with some simple >instruction. On the other extreme (improbably large effort) humans could do >even better if there was, say, an anti-computer opening literature to rival our >present opening literature. From your observations, have humans been able to >improve against computers by much? Do GMs _care_ to improve against computers? > >Walter I think that GMs do care _now_. 5 years ago, computers were a novelty. Now with the advent of internet chess, computers are a fact of life, and some GM players are learning how to play against them and being fairly successful at doing so.
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.