Author: Ernst A. Heinz
Date: 07:05:11 10/10/98
Go up one level in this thread
On October 09, 1998 at 22:44:27, Robert Hyatt wrote: >On October 09, 1998 at 22:29:18, Roberto Waldteufel wrote: > >> >>On October 09, 1998 at 20:21:50, Robert Hyatt wrote: >> >>>On October 09, 1998 at 18:05:18, Bruce Moreland wrote: >>> >>>> >>>>On October 09, 1998 at 17:33:07, James T. Walker wrote: >>>> >>>>>Is there a practical use for tablebases like the KBKN or KBKB etc. ? Am I >>>>>making a mistake by not using all the available tablebases ? >>>>>Jim Walker >>>> >>>>Those are probably not important. >>>> >>>>If you use something like KP vs K, make sure you have KQ vs K, or you might >>>>experience strange behavior as the program might forget that it should promote. >>>> >>>>bruce >>> >>> >>>Crafty won't even probe things like KB vs KN, it has what Ernst would call >>>an "interior node recognizer" that classifies such positions as draw, >>>without needing to do a probe of any kind. So if using Crafty, you can >>>delete those..., although you would need them to construct other >>>databases... >> >>Hi Bob, >> >>These are certainly drawn for practcal purposes, but mate positions do exist, so >>classifying these as draws without any testing for this could on rare occasions >>overlook a silly mate in one. Do you test for this in Crafty? >> >>Roberto > > >nope... KB vs KB == 0.00 in crafty, period... >ditto for KN vs KN, KNN vs K, KNN vs KN, etc.. Yes exactly -- Bob, I call these "interior-node recognizers" (albeit primitive ones) if they get called at interior nodes and not only in the evaluation function. But my implementation framework as detailed in the latest ICCA Journal 21(3) allows me to check for all the rare exceptions in the above mentioned endgames without any loss of efficiency. "DarkThought" does this kind of checking and accordingly *correct* interior-node recognition in many other cases, too, even if only imprecise knowledge may apply. =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.