Fail High Reductions (Extended Abstract) (1996)
| Venue: | Advances in Computer Chess 8 |
BibTeX
@INPROCEEDINGS{Feldmann96failhigh,
author = {Rainer Feldmann},
title = {Fail High Reductions (Extended Abstract)},
booktitle = {Advances in Computer Chess 8},
year = {1996},
pages = {111--127},
publisher = {Universiteit}
}
Share
 |
 |
 |
 |
||
OpenURL
Abstract
Fail High Reductions (FHR) is a new method to guide the search in game trees in a very selective manner. The main idea is that the search algorithm should not spend too much effort searching subtrees, which look too good to become part of the principal variation even at their roots. More precisely, a fail high node is a node v with a search window [ff; fi] at which a static evaluation function e produces a cutoff. The FHR-algorithm reduces the search depths at these fail high nodes thus searching their subtrees with less effort. FHR is domain independent in the sense that it only uses a static evaluation function, which always is assumed to exist, and the search windows to guide the search. In this paper we describe the incorporation of FHR in our chess program Zugzwang. Three different tests have been conducted to compare Zugzwang with and without FHR: First, we compare the results obtained from running the algorithm on the positions of the WinAtChess test suite. Second, we look at th...
Keyphrases
fail high reduction extended abstract static evaluation function different test selective manner new method winatchess test suite fail high node chess program zugzwang fhr-algorithm reduces much effort main idea principal variation game tree high node search algorithm search window ff search depth
