Author: Dann Corbit
Date: 16:33:39 05/30/02
Go up one level in this thread
On May 30, 2002 at 17:59:47, Miguel A. Ballicora wrote: >On May 30, 2002 at 17:42:19, Dieter Buerssner wrote: > >>On May 30, 2002 at 11:00:23, Miguel A. Ballicora wrote: >> >>>The ones that are out of real life are the tablebases (mate oriented) that just >>>ignore the The Rules of Chess. Note that the Thompson EGTBs do not have this >>>flaw. >> >>I wonder, whether this is true in general. Assume a 6-men TB (Thompson has >>generated many, one can look them up). Now you have the DTC, and you can judge >>this in the program. But say, the conversion is to some long mate in 5-men TBs >>(say KBBKN), that is a forced mate, but not in 50 moves without capture. Will >>the Thompson TBs show a draw score in the 6-men TB then? 6-men TBs I checked >>online, show happily very long distances to conversion. > >I assume that it was taken care of but I do not know. If the EGTBs are build >from the simplest ones to the more complex and it is stored "Draw" when there is >a mate that it is in more than 50 there should be no problem, or at least if it >says that it is more that 50 to return draw when building the more complex >table. I assumed that the DTC were built in this way, otherwise they may have >similar flaws. > >I think that the best EGTBS should be a combination of both methods, ie. the >information of both tables together in the same entry. I guess that they should >not be 2x bigger since most of the positions the DTM are ok and only that >information could be stored. > >>Such a scenario may also be possible in 5-men TBs only (but I am not sure, if >>such cases exist). Assume KPPKP, and after few moves 2 pawns promote to KQPKQ. >> >>Of course, it should be possible, to solve this problem with DTC tables, but I >>don't know, if Thompson tables take this into account, when the long line >>without the capture/pawn move is after the first conversion. > >I just started to code some elements for building my own EGTBs, (for the fun of >it). If I ever reach the stage of building the 5's I will really consider this >point. I propose a new rule: If the distance is 50+ moves to mate against a human, it is a draw for the human, even if he would be checkmated at move 51. No sense making them always shake hands or giving an annoying twit a way to force people to play on and on. If the distance is 50+ moves to mate against a computer, a tablebase checkmate with any finite number of plies is a win. After all, a computer will not tire out because he has to make 100 moves after a long chess game.
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.