Author: Ron Langeveld
Date: 15:03:23 06/25/02
Go up one level in this thread
On June 25, 2002 at 16:16:13, Uri Blass wrote: >On June 25, 2002 at 15:11:18, Ron Langeveld wrote: > >>On June 24, 2002 at 14:19:45, Joshua Lee wrote: >> >>>Is the reason for losses between fairly close programs >>>1. Book >>>2. The first Move out of Book >>>or something else? >>>I have looked at a few SSDF games and it looks to me that #2 seems to be a huge >>>problem. While it may take 60-70 moves or more but it appears to me that the >>>Game's aren't lost amongst the best in the Ending or MiddleGame As much as the >>>first Couple of moves out of book. >> >>Absolutely true. Between top engines the book determines the outcome of any >>comp-match for at least 70% (my wild guess). > >I do not believe in it. > >I expect part of the top engines with a very small book(only the first 3 moves) >to do more than 30% against the same program with the commercial book at least >in long time control because i believe that top engines can find a lot of book >moves by themselves and it is possible to avoid the same loss twice by changing >the opening in the first 3 moves. > >My opinion is that books are not very important. > >They are not important for weak engines because they often cannot take advantage >of better position and they are not important for the strong engines because the >strong engines can often find the right book moves by themselves and by choosing >the right book moves of the first 3 plies it is possible to get positions when >they can find the right book moves by themselves. > >There are positions when there are deep traps that almost every engine can fall >into them but it is possible to avoid them by smart choice of the first opening >moves(for example 1.d4 and 2.Bf4 when white can continue in many ways later and >it is practically impossible to plan something for most options or most engines) > >Uri Uri, Basically, I think that you are right. At least I believe that your theory with the 3 move book cannot easily be refuted in a test. The whole idea is very plausible. However, I do think that you are missing an important aspect here. One that I failed to mention earlier, and that is that currently all top programs use books that are quiet big, not as big as i.e. Powerbook, but big nevertheless, with many many lines going beyond move 20. It is not only my believe, but also my experience that a lot of these lines are "lost". Not lost in an absolute true verdict, but lost for the engine that is using it. Next time around it will avoid that line, only to stumble on another lost line. This is what we call a poor openingbook; they go so deep that any top engine can have major problems in trying to recuperate. Regards, Ron
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.