Computer Chess Club Archives


Search

Terms

Messages

Subject: a better idea of generating a detailed analysis tree for a position

Author: blass uri

Date: 22:06:31 04/01/98


fritz5 has a function of correspondence analysis such that it generates
a tree (every move in the tree is found by the fritz5 originial
program).
the method of generating the tree should be improved.

when we generate a tree we have to solve the following questions:
1)what is the next position that should be analysed?
2)what is the time that the computer will analyse it?

we should give every position in the tree a number and  to choose the
position with the biggest number(if this position was analysed before
we should compute the next best move in the position)

the number should be a function of some factors
1)d-the depth of the positions(number of moves after the root)
2)e-the evaluation function of the position
3)f-the number of forced moves that are in the moves before
the position(a move is forced if the difference between it and
the next best is at least a pawn)
the position that will be analysed is the position such that e+f-d
is maximal from the positions that analysing them can change
the main line
we sometimes should repeat analysing the main line for longer times.

it is interesting to do competition between different trees based on the
same program.



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.