Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Solving chess

Author: Janosch Zwerensky

Date: 09:24:10 08/02/05

Go up one level in this thread



>These techniques each reduce the average number of branches of the game tree by
>2

Alpha-Beta with perfect move ordering reduces the number of branches visited
during a brute-force search without hashtables to ply n to the square-root of
the corresponding number for a simple minimax search of equal depth. However, it
certainly does not necessarily lower the number of nodes visited during the
search by an equal amount; for an (admittedly trivial) example, in a game of
branching factor one, alpha-beta does nothing to reduce the nodecount of the
search. Moreover, even with a nontrivial game (i.e. a game with average
branching factor much larger than one) I don't see alpha-beta reducing the
number of branches visited during an exhaustive game tree search to the square
root of the size of the corresponding minimax tree when duplicate branches of
the gametree are removed by hashing. A simple game where the effect of hashing
on a minimax vs. alpha-beta comparison is quite obvious might for example be a
nim-like game with one heap of straws and a limit on the number of straws that
each player is allowed to remove on their turn. Alpha-beta with perfect move
ordering will in this case dramatically reduce the size of the unhashed game
tree, but do very little to decrease the size of the minimax game tree with
hashing.
Of course, all of this does not logically rule out the possibility that an
alpha-beta searcher need only search a game-tree of tractable size to solve
chess because it is logically possible that the starting position is a mate in
20 for white, in which case the solver can prune all subtrees of the chess game
tree which are worse than mate in 20, immediately removing probably a larger
number of positions than would be suggested by computing the number of positions
that can happen in a legal game of chess and taking the square root of that
number.
As for quantum computers, I doubt they will be of much help in solving chess
even in the case that practical quantum computers will in fact be built someday.
In this respect I am skeptical mostly because I do not see how to go from
unordered database searches, at which indeed quantum computers achieve a
quadratic increase in performance versus classical machines, to game tree
searches, which simply are a different kind of thing.

Regards,
Janosch



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.