Author: Peter Kappler
Date: 17:21:02 01/26/00
Go up one level in this thread
On January 26, 2000 at 20:08:04, Dann Corbit wrote: >On January 26, 2000 at 19:59:43, Dave Gomboc wrote: >[snip] >>>>Got an article that shows that the Elo curve flattens out with increased depth? >>> >>>Darn. I knew someone would ask that! I just read it somewhere, but I will have >>>to go and look for it now. >>>:-( >> >>While you're looking, beware of theoretical arguments that don't have the same >>conditions as the actual topic at hand. Every time someone attempts to support >>this argument using Nau, 1983(?) I shake my head. > >I got an email back from Nau. He said (among other things): "I doubt that the >work is relevant for chess-playing in any practical sense. I've only seen one >pathological game that people might actually sit down and play, namely Pearl's >game as described in my 1982 AIJ paper." > >>IMO, in this domain, experimental results are of more value. > >That's a good idea. Do we have a single program for which we have accurate data >for at least 4 different CPU speeds? By accurate data, I mean a huge number of >games under controlled conditions. By single program, I mean a single version >of a single program. Why not just take Crafty, and modify it to do fixed-depth searches with no extensions? Then play a bunch of Crafty vs Crafty matches with one machine searching n plies, and the other searching n+1 plies. My prediction is that the matches results will get less lopsided as 'n' increases. In fact, I'll be shocked if this isn't the case. If nobody else volunteers, then I'll conduct the experiment with Grok (but not until the ICC tournament is finished). :-) --Peter
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.