Author: Amir Ban
Date: 05:37:44 06/18/98
Go up one level in this thread
On June 18, 1998 at 07:00:17, Enrique Irazoqui wrote: >On June 18, 1998 at 03:58:08, Ernst Walet wrote: > >>All in all it is quite clear to me why Chessbase took so long to produce their >>auto232 player, they had to figure out how to kick Rebel's butt! > >This is false. Frizt 5 does not beat Rebel 9 in a different way than it beats >any other program. > >When I said that Rebel 9 did not learn in its SSDF match against Fritz 5, I got >mixed up with the Rebel 8 - Fritz 5 match. Rebel 8 played 13 losing doubles. >Rebel 9 played no doubles at all, as anyone can check. So all this discussion >about brain surgery, Rebel 9's learner being upset in that match, etc., makes no >sense at all because it's all based on this mistake made by yours truly. So much >for relying on memory when talking about facts :(((((((((( > >My apologies, Enrique That's still not quite right. The Rebel 8 match had 7 identical draws, which accounted for almost half the points it scored (8.5). I took interest in that match when it was debated and posted what I found here. I found that Fritz scored 79% in the match, while if doubles were discounted it would score "only" 77%. Big difference. This doesn't mean that Rebel could not become a victim of the Fritz learner, but the facts show that it didn't, this time. We now know that the Fritz learner will seek to repeat a draw. That this is risky strategy was shown in this match. A learner should try to repeat wins only, and prehaps should avoid not only past losses but past draws. Amir
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.