Author: Ernst A. Heinz
Date: 15:18:31 05/24/00
Go up one level in this thread
Hi Bob, >The idea is ok, but I don't like the concept of playing program X vs itself >with different depths. Your conclusion can easily be right for Fritz, but >wrong for other programs... This is certainly true -- and I do not claim to have shown anything else than just ``the existence of diminishing returns for additional search in computer chess self-play with 95% statistical confidence, exemplified by the program "Fritz6"''. No more, no less. As for self-play with depth handicap, it is a generally acknowledged technique to study the relationship in question. Several other researchers like Ken Thompson, Hans Berliner, Murray Campbell, Jonathan Schaeffer, etc. have used it for this purpose. But direct self-play represents only one half of my ongoing studies in this area. :-) > It would be hard to draw conclusions based on testing only > one program that is known to be very fast but not very 'smart'. Right, we cannot draw any general conclusions from this. But it certainly is a proof of existence -- no more, no less as I said. Moreover, I am not so sure that "Fritz 6" is far less smarter than other programs. According to personal communication with Frans Morsch, "Fritz 6" does full tip-node evaluations and hardly any root-node preprocessing. Given the relative speed decrease in NPS of "Fritz 6" as compared with "Fritz 5", I have no reason to doubt Frans' statement. =Ernst=
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.