Author: Vincent Diepeveen
Date: 06:05:25 01/30/02
Go up one level in this thread
On January 30, 2002 at 04:29:40, Tony Werten wrote: >On January 30, 2002 at 03:41:38, Les Fernandez wrote: > >>On January 30, 2002 at 03:14:40, Tony Werten wrote: >> >>>On January 28, 2002 at 17:48:48, Les Fernandez wrote: >>> >>>>Just curious to know what is the best being done now for storing a chess >>>>position. In the worse case scenario (castling right exists) it takes me 162 >>>>bits to store 32 pieces, color, location, side to move, castling, enpassant, >>>>promotion, ce and pv. Now if castling rights do not exist then the worse case >>>>scenario for the above is 81 bits. Much further reduction in bits/position is >>>>possible but at the moment I am interested in the above. >>> >>>Hi Les, >>> >>>I read your paper but I don't get it. I understand the permutation stuff etc, >>>but how does it work when you don't know wich piece is going to checkmate ? >>> >>>Suppose I have a KBNk endgame. How am I going the permutate. For king, bishop or >>>knight moves ? In your capdata you might know wich piece is going to move but in >>>search you don't. >>> >> >>Hi Tony actually all we are interested to know is where the pieces are. Once we >>have that we can look for a match using the above metioned technique. Any other >>questions let me know. > >I see. I like the idea more and more. Not for endgames but for opening. Put some >typical positions that a computer should avoid (or reach ) into a small >database. You'd require no_of_pieces lookups, (not too expensive for say the >first 6 ply ) or 1 lookup but then you need 7 ply. > >Nice idea. Unfortunately i don't need new ideas, I need time to test the old >ones. it is not a new idea even. they invented new terms for something which already existed in several programs for egtbs (in order to speedup the egtb cache). >Tony > >PS Ed said something about historytables that is vaguely related to this. He >doesn't use piece,fromsq,tosq in them but piece,tosq since the good move was >move the piece to tosq, independently of where it was coming from. history heuristic is something different from wasting some ram to 'known' mates somewhere in the huge gamespace of chess. >>Les >>>So just permutate for all moves ? I think I'd rather search a ply deeper, but >>>close to root it might not matter that much although it would be quite a >>>computational and memory stressed effort. >>> >>>Tony >>> >>>> >>>>Thanks in advance, >>>> >>>>Les
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.