Author: Uri Blass
Date: 14:50:24 12/28/04
Go up one level in this thread
On December 28, 2004 at 17:14:44, Kurt Utzinger wrote: >On December 28, 2004 at 16:54:44, Uri Blass wrote: > >>On December 28, 2004 at 16:16:08, Kurt Utzinger wrote: >> >>>On December 28, 2004 at 16:12:18, Uri Blass wrote: >>> >>>>On December 28, 2004 at 15:26:16, Kurt Utzinger wrote: >>>> >>>>>Matches at 40’/40 + 40’/40 +40’ time control >>>>>Junior9-GUI, ponder=off, 3-/4-men EGTB >>>>>own books, no book learning, no learning >>>>>on 4 Athlons 1.3/64 MB hash for all engines >>>>>First defeat of Junior9 vs Shredder8 in this test. >>>>>http://www.utzingerk.com/jun9_test.htm >>>>>Kurt >>>> >>>>If you use books and use no learning there is a risk for double games. >>>> >>>>Uri >>> >>> Hi Uri >>> Yes, this can happen. But such games will be cancelled >>> and replaced by new ones (some more work to do of ...) >>> Kurt >> >>I can imagine that this method may allow no testing in some cases. >> >>It is possible that 2 programs will repeat always the same game if you do not >>allow learning. >> >>The programs may have a big book but only one strategy is enabled for them in >>case of no learning and chooising a different book move may happen only after >>learning from previous games. >> >>Uri > > I have never seen two programs to always repeat the > same game. The wide range of opening lines of the > opponent will not "allow" this. > Kurt The fact that a program have in the opening book both 1.e4 1.d4 and 1.c4 does not mean that it has to choose more than one of them with no learning. It is possible that a program choose always 1.e4 in case of no learning and the same for the rest of the positions in the book so it will not lose the same game twice with learning because it may learn not to repeat the line that it lost but even after learning it may still have only one possible move that it can play in every position(the difference after a loss may be only changing 100% probability for 1.e4 e5 to 100% probability for 1.e4 c5 GM's do not choose randomely moves between the theory moves that they remember when they play and I see no reason that a program needs to behave in a different way. 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.