Author: Mogens Larsen
Date: 09:32:15 05/29/01
Go up one level in this thread
On May 28, 2001 at 23:33:31, Dann Corbit wrote: >Not at all. A program without an opening book (or worse yet -- a bad opening >book) will play like an idiot compared to one with an excellent opening book. >That is the major difference between the professional programs and the amateur >ones. However, the amateur engines are getting much better books, allowing them >to play a lot smarter. They won't play like idiots because they don't actually play the opening moves. The after book position might not be advantageous, but that is something different. Tablebases have the same flaw. No moves are being generated by the engine. Essentially it doesn't play chess by definition. I think that is a cognitive problem. It signals the intention of transporting an algorithmic structure from start to finish before it's terminated externally by making the journey shorter (bigger books and endgame tables) instead of improving the survivability of the structure itself. >Classical chess openings have been debugged by thousands of supercomputers (the >human brain). As time goes on, we figure out what works and what does not. >Just as a good human player *must* add many sound openings to his repertoire to >play good chess, so also the computer must do the same. Can you imagine two >players with GM ability, but one who has never studied openings and the other >who has studied them for a decade? How do you suppose it will turn out for the >neophyte with equivalent talent but not equivalent knowlege. That's a very good argument for testing without opening books, ie. to nullify external input outside engine control. The scientific interest in watching computers construct their own opening books based on experience instead of being ruled by human perception of absolute chess knowedge. The GM players example has nothing to do with this question, because it's a random assumption not related to the experiment. It's far from nonsensical to test without books, but it requires the ability from both engines to store and use games already played in order to diverge in a sensible way to avoid (and maybe also encourage) repetition. Adding an opening phase algorithm of some sort to aid diversity won't necessarily weaken the engine as you claim. But it's not a pretty solution. Regards, Mogens
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.