TalkChess.com

Laszlo Gaspar · Joined: 09 Mar 2006 Posts: 52 Location: Budapest, Hungary

Hello guys,

The theory behind this idea is well worked out. It is called: Alpha-Beta Conspiracy Search by David A. McAllester and Deniz Yuret. They measured the stability of the evaluation as the number of leaf evaluations which are necessary to change in order to change the outcome of the search. They introduced two depth parameters for both sides which mean the number of moves below a certain node in the tree which already reached the established alpha - delta (usually a half pawn value) value for the side to move. When both sides reach their 'depth' the search is aborted.

I implemented it and it did something just it was very resource hungry. When you generate the moves you also have to evaluate them with at least a quiescent search knowing instantly it is good or bad. And the good moves are counted in the tree. It was a very interesting experiment but I realized that it is simpler just extending check by one ply... Smile

Good luck! Smile

_________________
Regards,
László