Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: EinStein wuerfelt nicht - Challenge Game

Author: Ingo Althofer

Date: 22:52:18 03/05/05

Go up one level in this thread


On March 05, 2005 at 19:47:18, Dieter Buerssner wrote:
>... Does anybody have pointers, how
> such games, that include chance and strategy, are programmed?

Game tree search is possible. However, there are (almost)
no alfa-beta cuts, due to the levels with random action.

Interestingly, a very simple Monte Carlo approach gives
already good results: play many completely random games
for mthe current position and select that move with the
best score.
When you make tree search, the leaves may be evaluated
by such simple Monte Carlo runs.

Stefan Schwarz designed another (quick) evaluation
function: For the current position 2 values are computed:
(i) the expected number of moves for player X to reach the goal,
assuming that there are no stones of player Y on the board.
(ii) vice versa the expected number of moves for player Y
to reach the goal, assuming that there are no stones of player X
on the board.
The difference of these terms is the position's value.


>>Perhaps my next move will come
>>during one of the advertisement brakes.
>                                 ^^^^^^
>
>Good typo in the context of racing.

Yes, and in contrast to the opinion of the TV guy
the race was boring. When I would have known before, I would
have kept sleeping.

Ingo.



This page took 0.02 seconds to execute

Last modified: Thu, 07 Jul 11 08:48:38 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.