Author: KarinsDad
Date: 15:34:53 06/09/99
Go up one level in this thread
On June 09, 1999 at 17:47:05, blass uri wrote: > >On June 09, 1999 at 16:39:42, KarinsDad wrote: > > ><snipped> > And why would black as a perfect tablebase >>program play for a draw when it would consider 1.b3 a slightly inferior move for >>it's opponent? > >perfect tablebases do not know what is slightly. >The only knowledge is for the result for every legal position. > >perfect tablebases can tell you to do stupid moves(to give a pawn in KRP vs KR >and to lose the practical chances for a win). > >If the starting position is a draw then only losing moves are considered to be >inferior by the tablebases. > >I understand that you have a different definition of perfect tablebases so what >is your definition? My definition can best be shown by example: 7k\Q7\7K\8\8\8\8\8 White could win with Qg7++, Qh7++, Qa8++, or Qb8++. Qf2 and Qf7 are stalemates. Other moves can lead to a draw or a win, but not a loss. Therefore, the perfect moves are the first 4 that I listed. Although other moves may lead to a win, they are not perfect in the sense that they take more moves and hence, they do not need to be in the tablebase (in regard to this position). Neither do the draw moves. Neither does this position if a perfect game does not lead to it. Neither do 3 out of 4 of these moves. The reason I place the minimal number of moves restriction is in order to minimize the possibilities. I do not use the current concept of a tablebase that returns the number of moves until a draw or a win with no concept of a better or worse move (the program makes that determination based on ply). I use the concept that some moves are perfect in the sense that they will lead to the best result in the given position (assuming that the player continues to make best moves). As can be seen by my example, there could be more than one perfect move in a position. Therefore, the perfect tablebase (there would be 2, one for white and one for black) would not have ALL legal positions. For a given side, it would only have those positions that can be achieved by it's opponent; assuming that the side using the tablebase is making perfect moves. So, this would drop the maximum number of positions for 60 moves each from approximately 40^120 to 40^60 positions in the table (ignoring transpositions, ignoring early draws and wins, assuming that you only need one perfect move for any given position in the table, and assuming an average of 40 legal moves per position). Obviously, the real number of positions in the table would be much smaller than 40^60. KarinsDad :)
This page took 0.01 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.