Game: Chess: Barney Pell Way
Implemented by David GLAUDE, January 2001.
This Zillions Rule File is dedicate to Barney Pell for it's innovative research in game learning.
I am only the translator into Zillions language of rules suggest by Barney Peel.
You can contact me David GLAUDE or browse my web site. Chess: Barney Pell Way, contain in one zrf files, three variants of Chess that can be used to evaluate a Metaplayer's evaluation function.
Two of those variants where suggested to me (David GLAUDE) by Barney Pell in order to estimate the Zillions-of-Games engine.
The last one is taken directly from the PhD work of Barney Pell.
The traditional game. Object: Checkmate the opponent's King.
Traditional game where Knight are replace by Super Knight.
A Super Knight move 3+1 rather than the 2+1 of a normal one.
This extended move capabilities also make the the Black Super Knight will always stay on Black square,
and White one stay on White.
Also, it is very difficult to move a Super Knight arround because you are almost always close to the edge.
Will Zillions find the real value of that piece compare to the original?
Traditional game where Knight promote when they reach the promotion raw.
Just like Pawn, the value of a PromotingKnight is having an increased value.
It is worth keeping the PromotingKnight until the end of the game, but will Zillions discover that?
Traditional game, but the board is surrounded by and extended zone only reachable by the ExtendedKnight.
Because that zone is reserved for the ExtendedKnight, this one can hide from capture and attack from behind.
Will Zillions find the way to benefit from that new powerfull weapon wich is the ExtendedKnight?
Download Chess: Barney Pell Way Now!
The PhD research of Barney Pell focused on developing programs that could play games directly from the rules, without human assistance. He called the idea Metagame, and implemented the first program in this new paradigm.
Metagame is the brainchild of Barney Pell, you can learn some more about him on he's web page.
Jay Scott on he's web site dedicated a page on Machine Learning in Games and one on the Metagame project.
The pages from that web site and the extensive documentation show us that there is more than one attempt to make machine learn to play new games, describe by rules or move generator.