Author: Mike S.
Date: 14:06:30 12/12/00
Go up one level in this thread
On December 12, 2000 at 13:30:47, Amicitia Stone wrote: (...) >All I'm saying my man is that your question "which is best" cannot be answered >so easily without a massive debate. Indeed. It starts with the definition of "best book". Some will say, the best book for a program is the one, with which it achieves the most success in comp-comp games. In other words, the book that optimizes the chances of the program, to be ranked high in comp rating lists. This would be the rating list thinking, applied to the book discussion. The more the success of a program comes from the quality of it's book, the less I'm impressed of the engine. I have written an article for CSS once, based on an experiment I made to test books. I tried to determine 3 parameters (for the tournament setting): 1. average variation depth in plies 2. variability (measured by counting the doubles which occured, while attempting to produce 30 different variations) 3. evaluation of the engine at the end of the book variation (average) My method was somewhat limited, because I have tested only one book each, by self-play (which was quite an effort anyway); I did not have autoplay equipment and the engine match options were not yet developed then IIRC. Btw. I gave several suggestions in that article, how the method could be developed further etc... My impression from the response (which was non-existant AFAIK) is, that hardly someone will make the effort of in-depth book testing, determining book quality parameters (not necessarily those I suggested) and comparing opening books. What is done are program matches with various books occasionally, and the results are 8-12 or 11-9 or such dead numbers. I have never seen statistics which would tell for example, how many of the wins came from the books, and how many from the engine fight alone more or less. Currently my general impression is, that there is by far not that much interest in details among the computerchess fans, as it may seem by the number of postings, games and programs etc. Regards, M.Scheidl
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.