Author: Mark Young
Date: 11:39:31 06/15/98
Go up one level in this thread
On June 15, 1998 at 13:54:26, Vincent Diepeveen wrote: >No doubt that Fritz5 book kills again Rebel9 book, statistics are clear in >this from SSDF. I didn't receive the special SSDF fritz5 book that kills >Rebel9 there, so i could not test that book at auto232 player against >rebel9, too bad. > >Yet it is clear that just looking to how the games go and >how the score is of Diep first move out of book that this rebel9 book kills >the genius 5 book. > >1 game diep comes out of book with -0.24 , -0.44 , -0.58 then >we get to -0.78, -0,84 and further ALL games are less then -1.xx > >Out of 14 games this make just 2 games it comes out of book less than >half a pawn down. 3 games almost a pawn down, and the rest more >than a pawn less down. > >So average diep comes out of book with 1 pawn down. Not what i >call an equal chance. > >We cannot blame SSDF, we just need some wide books, >this is clear to see. No i don't think Rebel9 >proves anything by killing in a prepared way the genius5 book, nor >nimzo98 by killing the rebel8 book, nor fritz5 killing the rebel9 and all >other books. > >Nimzo98 book simply kills the rebel8 book terrible. > >Perhaps Ed wants to produce some statistics rebel8 book+rebel9 or rebel8 >engine against nimzo98. > >Interesting would be nimzo98 with 3.5 book against rebel8 book + rebel9 or 8 >engine. Same for nimzo 3.5+nimzo98 book against rebel8+rebel8/9 engine. > >I have my suspicions what the results will be. the newer book simply always >wins. Not the engine. > >Mchesspro i don't have too much games from. Very impopular program in >Europe, but i hear it does the same. > It does. >This all pleads of course for really wide books, so programs need to be >allround players then, unless you just want your program to win for a few >months until a new book comes out of the competitor! I agree. I have made a wide and deep book and for Fritz 5. I find the programs is much more interesting to play also. The draw back is disk space. The book is 750 mb and growing. In the future maybe programs will not come with a premade book. Just games and download sites for you to make your own book out of good quality games. This way all programs can play with the same book when testing. Its just a dream. The programmers will never go for it. >So engine then needs to be improved then! > >Note that Grandmasters when playing other grandmasters are not any better. >Simply preparing for their opponents certain lines, yet they are smart and >clever but don't have the 100% sureness that thousands of testgames give >that you beat your opponent, because very likely this opponent doesn't >play the 100% same. > >Vincent
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.