Author: Volker Böhm
Date: 02:36:45 05/13/04
Go up one level in this thread
On May 13, 2004 at 05:06:03, Daniel Shawul wrote: >On May 13, 2004 at 04:34:36, Volker Böhm wrote: > >>Hi, >> >>if (r_beta == 4,75) and the real score is 10,00 the fail-soft score can be >>anything between 4,75 and 10,00. >> >>Any returned value >= beta will immediately result in a cutoff and return the >>current best score. The alphabeta algorithm will not check if there are other >>moves with higher score. >> >>Thus in your example at some ply there is a move tried first that returned 4,75 >>and the best move which would return 10,00 is never checked. > > That's is a fine explanation. But whatever margin i use >2.0 , 2.33, 3.45 etc the score i get is always (beta + margin). >Does this mean there is a (beta + margin) score somewhere in the tree?? If it does happen allways it is not verry probable that the score is somewhere in the tree. Check if you use alpha or beta as result value anywhere! Perhaps in lazy eval? Perhaps when foreward-pruning? Perhaps when pruning in q-search? Perhaps when no good enough move could be found (all moves < alpha) ... > >> >>Greetings Volker
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.