Author: Mike S.
Date: 21:27:58 08/26/02
Go up one level in this thread
On August 26, 2002 at 23:34:15, Will Singleton wrote: >It's common wisdom that a good book will increase a program's strength, and I >have no doubt about it. But, what about the typical amateur book, which is >usually created from pgn files, without a great deal of hand tuning? If it is generated from (recent) GM *practise* only, it may fail at the first of the following purposes: I think that good opening books have two major effects which are important (too, except the book learning): 1. avoid deep opening traps (for games against other books which might try such traps, or "killer variants", which do not appear in GM practise) 2. save time. A deep book variant can probably save up to 50% of the thinking time per game (as long as there are book replies to the opponent's moves of course) While an engine can probably "survive" against an average book, or against a book set to a narrow repertoire, it will most probably have big trouble against a gambit book or against a book (-setting) which favours traps. An example (imagine Black has no book): 1.e4 Nf6 2.e5 Nd5 3.Nf3 d6 4.Bc4 Nb6 5.Bxf7+ Kxf7 6.Ng5+ Kg8 7.Qf3 Qe8 8.e6 g6 9.Nf7 Qc6 10.Nh6+ Kg7 11.Qf7+ Kxh6 12.d3+ Kh5 13.g4+ Kh4 14.Rg1 Qd5 15.Nd2 Bh6 16.Nf3+ Qxf3 17.Qxf3 Bxc1 18.g5 Bxe6 19.Qe4+ Kh5 20.Qxe6 1-0 I'm sure it would be easy to find engines/books which are not prepared for 5.Bxf7+?, while good books know the line. (Btw. 6...Kg6 7.Qf3 Kxg5?? is suicidal too.) Some engines are able to avoid the wrong moves by themselves, but I'm sure not every engine, and not at any time control. Also, there are more difficult traps similar to that. Some are known from (old) book killer variants. >(...) So, could a crude book, created >from pgn files, significantly help the typical amateur program? I think that most often book options will be set to a more narrow tournament repertoire, when people let play tournaments, create rating list etc. While I thought until now (too) that this is an obvious choice, I'm beginning to doubt if this is good... Anyway, a normal "generated" GM book will help much, at least save time, as long as main variations are played, or lines which have appeared often in GM chess. >I felt I should conduct a small test involving my program, Amateur. I suggest testing it with risky, or unusual trap-like variations, some of which may never have appeared in GM chess, or are only to find in gambit books and the like. Also, comprehensive opening books should mention many of them in the side variations or remarks. Some chess programs also have gambit- or "sharp" books which should try such variants. You could take a look what happens if your book&engine runs into such variants, i.e. if the engine makes the "??" move when out of book and falls into the trap. IOW, "generated" books can probably be improved by adding theory which isn't in the GM practise. Unfortunately it sounds like much work, move by move... :o) 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.