Author: Tony Werten
Date: 06:49:13 04/08/01
Go up one level in this thread
On April 07, 2001 at 11:18:34, Urban Koistinen wrote: >I have written down a algorithm for computing endgame tablebases that should be >about 10 times quicker than the Nalimov algorithm and requires much less ram. >It is similar to the Arlazarov&Futer algorithm of 1979 but is more general as it >does not require a pawn. >It might be too technical for most here, but if anyone has questions I will try >to answer them. I don't claim I completely understand it, but here are some remarks. - It better be smaller than Nalimov tables. Nalimov uses 16 bits to give distance to mate, you use 1 bit to show wether it's a win. - t100 black lose with g50=100 t99 white win with g50=99 Where's t100 white win with g50=100 ? ie Why do you assume white has made the last move that sets the counter back to 100; it could have been black. - combining tables t0 to t98 with 6 bits/position is the real challenge I think. - Using the tables: How do you make sure you make any progress? ie From current position I have 2 moves. One checkmate in 10, one in 20. How do you see the difference ? Both moves will have counter g50-1, both will be a win. cheers, Tony > >The text is available at www.abc.se/~m10051/eg.txt >It is ok to copy it verbatim.
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.