Author: Terry McCracken
Date: 10:53:29 04/17/02
Go up one level in this thread
On April 17, 2002 at 11:24:38, Roy Eassa wrote: >On April 17, 2002 at 11:03:57, Terry McCracken wrote: > >>On April 17, 2002 at 10:13:33, Roy Eassa wrote: >> >>>Terry, it seems that in many tactical situations today's top programs running on >>>fast PCs may be Elo 3000+. That's why it has become so critical for GMs to >>>adapt their play -- those who don't will lose more games than they win. But I >>>think GMs CAN still improve their anti-computer techniques by a very wide margin >>>(that's not to say that every GM will). >>> >>>The science of anti-computer techniques is still in its infancy, at least >>>compared to the traditional chess techniques that have been refined continuously >>>for centuries. >> >>I think they can be 3000+ in short range tactics, as opposed to long range >>tactics. >> >>This strength manifests itself especially in fast blitz. >> >>Terry > > >The _width_ of the search is what impressed me most. Depth humans can do, but >the combination of good depth and impressive _width_ at a given depth is where >computers shine most, IMHO. They will find that one tiny, hard-to-find line >that refutes a whole gigantic branch of the tree, nearly every time. > >But again, I state my belief that at least some humans (not MOST, but SOME) will >improve dramatically at discovering and putting into practice techniques of >avoiding this ultra-sharp razor. There I concur, that the width of search is better for computers as a rule, for tactics anyways, but only short to mid-range. But that can be seriously compromised by techniques you're proposing. Computers are also better at finding mates at least mates that are filled with checks and captures as width isn't needed and they can calculate this ...well with the advantages of speed of a computer. But width can be a weakness too, as in holds back depth!;) Terry
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.