Author: Uri Blass
Date: 05:05:15 02/21/05
Go up one level in this thread
On February 21, 2005 at 06:30:57, Tony Werten wrote: >On February 21, 2005 at 04:12:55, Uri Blass wrote: > >>On February 21, 2005 at 03:54:03, Dieter Buerssner wrote: >> >>>On February 20, 2005 at 18:51:41, Dieter Buerssner wrote: >>> >>>>On February 20, 2005 at 18:40:34, Guido wrote: >>>> >>>>>You are right but if I remember well there is a mate in 127 and a loss in 127 in >>>>>kppkp and another score to add is illegal position. >>>>>So for kppkp it is necessary at least 9 bits. You could use the illegal position >>>>>score meaning loss in 127, as illegal positions are not possible in the game. >>>> >>>>Hi Guido! IIRC, 8 bits was just enough for kppkp and all 256 numbers were used. >>>>kppkp wtm has mate in 127, btm has lost in 127, together with draw and broken >>>>positions, this exactly fits the 256 states possible in one byte. >>> >>>Oops, you were right, it seems. I forgot about loss in zero. Sorry for my >>>useless post. BTW. the max positions here give fast Q-endgame, which is draw due >>>to 50 moves rule. >>> >>>Dieter >> >>I do not understand. >> >>Do you say that computers with the old tablebases evaluate the max position as >a draw instead of mate and that they are right because of the 50 move rule? >> > >Nope, they evaluate the max position as a win instead of a draw and they are >wrong because of the 50 move rule :) > >Tony > >>Uri I do not understand. How they can evaluate it as win when the table is broken because of 257 values. Do they say mate in 0 instead of mate in 127? Note that from movei point of view mate in 0 can be considered by tablebase as broken position with no change in playing strength. I think that using tablebases to find if position is mate is very expensive way that is not needed. Uri
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.