Author: Bruce Moreland
Date: 15:36:11 08/24/00
Go up one level in this thread
On August 24, 2000 at 17:37:31, Hugh Cumper wrote: >Wanting to practice some simple endings I set up R+P V R and gave Fritz the >stronger side. I was surprised that it was giving me very easy draws, sometimes >repeating moves or else allowing the rook exchange with an obviously drawn pawn >ending. The reason was that the tablebase told it the position could not be won >so it regarded all moves (that didn't blunder the rook) as equivalent. When I >disabled the tablebase access it gave me a much harder time. Wouldn't it be >better to use the engine to find a constructive move in a case like this, just >using the tablebase to check both sides moves for outright errors? It may be >that anyone who loses such a simple ending would be too weak to get into one >against a computer but this is surely not true of R+B V R or B+B V N. >I believe other tablebase enabled programs do the same thing, unless anyone >knows different. Yes, this is an incomplete implementation. The proper thing to do is, once you are in the tables, probe the tables to see if the game is a win, and if so return the proper score. If not, you hit the heuristic eval function and return that. bruce
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.