Author: José Carlos
Date: 07:12:44 08/26/02
Go up one level in this thread
On August 26, 2002 at 10:02:53, Uri Blass wrote: >On August 26, 2002 at 09:50:47, José Carlos wrote: > >>On August 26, 2002 at 09:27:10, Steve Coladonato wrote: >> >>>On August 26, 2002 at 09:02:12, Matthias Gemuh wrote: >>> >>>> >>>> >>>>I (www.gemuh.de) am of the opinion that books should by law be limited to 14 >>>>moves. This is enough to (almost) finish development and let the engines take >>>>over with almost all fighters still available for the battle. >>>>Engines should not be helped (???) beyond this point. >>>> >>>>Matthias. >>> >>>Matthias, >>> >>>I tend to agree with you here. It seems that "chess programs" are more a >>>function of the "book" rather than the capability of the engine. Yet somewhere, >>>the engine has to go out of book so why couldn't that be after 14 moves and why >>>couldn't the SSDF provide a "standard" book for tournaments. Are they >>>evaluating the engine or the book's author? >>> >>>Of course having a "standard" book means nothing if, in fact, the book usage is >>>embedded in the engine as mentioned by other threads here. Why can't book calls >>>be a routine separate from the basic engine code. But I don't program engines >>>so I don't really know. It just seems that the engines performance is skewed >>>way too far towards the book it uses rather than it's analytical capabilities. >>> >>>Steve >> >> If you're only interested in analytical capabilities, a match with same book >>won't work, because pondering scheme, time management, asymetrical eval, etc >>will make the test worthless. If you only want to test analytical capabilities >>you'd better use a big test suite, IMO. >> >> José C. > >I do not think that there are good test suites so it is better to use games when >the program use exactly 180 seconds per move and no pondering. > >The only pronlem that cannot be solved by this is assymetical evaluation but >I already dislike these programs for analysis so we can ignore programs with >that problem. > >Uri I guess programs with assymetrical eval disable it on analyze mode (Crafty?), so they would still work on solving test suites. But anyway, using fixed time and no pondering would definetly be a better estimate for analyzing capabilities. Not perfect, though, because if I want to analyze games, I'd probably choose tiger for sharp positions, hiarcs for endgame or weak-pawns positions, etc. I mean, _games_ will not tell you how good is this or that program for this or that kind of position. José C.
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.