Author: James T. Walker
Date: 08:43:11 03/26/02
Go up one level in this thread
On March 26, 2002 at 10:56:07, Vincent Diepeveen wrote: >On March 26, 2002 at 10:24:38, James T. Walker wrote: > >James, > >book learning is very important when you have a very good >initially tournament book. > >Because you can keep repeat a winnign line then. If such >a strong tournament book kills completely automatic books, >then PGN generated books make no chance simply. > >You get results like 50-4 then or similar. > >So important in the above example is the fact that you >speak about 2 different books where the important features >of learning is REPEATING won lines and avoiding lost or drawn >lines. > >Usually this happens when 1 book is much better than the opponent. > >So retry your experiment using auto232 player i'd say. I see HUGE >score differences between using learning and without learning. ********** My experiment was with auto232 on 2 AMD 1.4G machines. As you know with Fritz GUI you cannot turn off Book Learning in auto232 mode. That is not the idea anyway since we know that if one program has no "learning" then the one with learning will find a "killer line" and repeat it forever. As far as I can see this avoiding "loseing lines" by "book learning" is the only value. So it puts everybody back on "even ground" again. *********** >The difference between both using learning is near to zero of course, >because there is not a single dude in history yet who could define >intelligence. > >So if there is 1 path to always beat the other guy, then this will >be found with book learning. It isn't actually *improving* a book >of course. > >Improving a book is a manual thing to do. > >The real cool thing from learning is for SSDF purposes of course. > > >>On March 26, 2002 at 09:36:41, Vincent Diepeveen wrote: >> >>>On March 25, 2002 at 13:35:33, James T. Walker wrote: >>> >>>at auto232 it is impossible to turn off booklearning in fritz. >>> >>>how did you turn it off? are you SURE you turned it off? >> >>Hello Vincent, >>I guess I was not very clear in my original post below. I did not turn off any >>book learning. As you say, it's impossible with Chessbase GUI. I simply >>cleared the book learning in one computer and let the other computer learn from >>1700 previous games. The purpose was to see if the learning function from 1700 >>previous games helped in the match vs the computer which had no previous >>learning experience. In my 200 games the score was 103-97 with the computer >>with no previous learning the winner. So it appears there is no advantage to >>book learning except to prevent one computer from finding an opponents weak >>opening and playing it over and over again to get wins. Of course this is very >>important. I was expecting the computer which had learned from 1700 previous >>games to actually use this info to get a better score vs the one without any >>previous learning. >>JIm >> >> >> >> >>>>I just did a quick test to see if there is any gain through book learning. I >>>>loaded One computer with Fritz 7 and let it learn from the 3 databases I have >>>>(more than 1700 games played by Fritz 7. In the other computer (both AMD 1.4G) >>>>I cleared the book learning in Fritz 7 and played 100 games at G/1 minute (for >>>>quick results of course). The final score: Fritz without previous book >>>>learning won by 52-48. >>>>Comments?/Conclusions?/Insults? >>>>Jim
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.