Author: Vincent Diepeveen
Date: 12:21:13 12/22/99
Go up one level in this thread
I do not know whether tiger uses other learning than book. If it does, it would not be new. Genius3 already did something (it never repeats its games). fritz3 did some hashtable learning which all can be cleaned by setting the level to a certain depth (say 10 ply). Personally i think that preventing to lose in the same way is very smart. double games are indeed not counted. This is something we both agree on for this match. I find it weird that SSDF accepted it from Rebel8 though. On December 22, 1999 at 15:08:29, blass uri wrote: >On December 22, 1999 at 10:27:41, Vincent Diepeveen wrote: > ><snipped> >>Well about the DIEP - Tiger match >> >>DIEP (single cpu version) with book without learning >>against Tiger without book but with learning. > >1)I did not know that tiger has learning without book. >I understood that tiger's learning is only avoiding repeating the same losing >line again. > >I understood that doubles will not be count for this reason. > >2)I guess that tiger is going to win the match because it is simply impossible >to be number 1 in the ssdf list only by a book. > >It is possible only if you do a lot of testing against computers and save the >games you won as part of your opening book but I do not believe that someone >does it and if someone will work hard to do it then it will be revealed when the >program is going to lose against weak opponents that it did not train against >but tiger did not lose matches against weak opponents in the ssdf games. > >Uri
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.