Author: Jaime Benito de Valle Ruiz
Date: 17:20:33 12/14/05
Go up one level in this thread
On December 14, 2005 at 15:49:44, Joseph Ciarrochi wrote: >hi folks > >(i hope you will tolerate my simple question. i am not a programer, but am >interested in how chess programs think...I'm a research psychologist) > >With humans, the best blitz players have the most knowledge. This is because >pattern recognition is much faster than deliberate search > >But i get the sense that it might be the opposite for computers? ARe the most >knowledgable programs (e.g., hiarcs) expected to be worse at blitz than ones >that are optimized for searchd depth? I have heard that the key to blitz is >simple search depth? > >Or have i got it all wrong. Is there indeed a trade off between search depth and >knowledge? > > > >best >Joseph Humans use extensive pattern recognition to play chess, whereas computers rely probably more on search, yes. However, your question appear to assume a dichotomy between search and knowledge, and unfortunately it is not that simple. There are uncountable factors to take into consideration, including the weights used in the program, the kind of position that is being played, the speed of the machine, the memory, the hash size... you name it. Change any of them, and you will get unpredictable results when you make the engine play against an oponent of a strength that can be compared to this engine's strength. Change a test position, and you'll get unexpected differences. It is so complex that nobody will be able tell you exactly how or why for sure. A good eval can compensate for the lack of depth search sometimes, but some other times, a bit of extra depth can improve the quality of the overall decision even over a better static evaluation. Too much search, too blind, can be bad; and the opposite is true too. Mathematicians will find more to study here than psychologists, if you ask me. But in any case, the best way of learning is trying yourself and arrive to your own conclusions, like most of us do here (including programmers). Regards, Jaime
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.