Author: Will Singleton
Date: 19:51:23 07/08/04
Go up one level in this thread
On July 08, 2004 at 21:58:34, Norm Pollock wrote: >What exactly does "book learning" do? > >And suppose multiple engines are all using the same book in a tournament, and >that book learning is on. Won't the book learning of engine A influence the >opening choice for engine B? > >Wouldn't it be much better for each engine to have its own personal copy of the >"shared" book? Not having written my own book-learning code (it has been crowded-out by more interesting pursuits), I'm not qualified to answer. So here's my take. Book learning can mean anything an author wants it to mean. You can write code to notice bad opening lines, and weed them out. You can notice good lines played by opponents, and add them in. You can alternatively adjust the weights of your book moves based on results, or based on subsequent evaluations, or based on reverse analysis performed after the fact. You should read Hyatt's ICCA article on the subject for one approach. To your question about multiple engines using the same book, I suppose you refer to the case where engines compete under a gui that uses its own book for both engines (that is, neither engine does anything until the end of book is reached). Then book-learning is not possible, since the engines are not using their own book and cannot modify it. However, position learning might be useful in such a case, though that's not your question. If you instead are referring to a situation in which multiple engines are using the same book, and each have their own local copies (gui not controlling things), then that would be the same situation as if they had books that were different. Normal book-learning would apply, and of course, depending on method, subsequent games would be affected.
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.