[ List Earliest Comments Only For Pages | Games | Rated Pages | Rated Games | Subjects of Discussion ]
Single Comment
Angels and Devils. Chess game where white has two Angels and black has two Devils. (10x8, Cells: 80) [All Comments] [Add Comment or Rating]
💡📝Greg Strong wrote on Sat, Jul 2, 2005 04:30 AM UTC:<p>I will post a ZRF shortly, for completeness. However, a large number of
tests have been run with ChessV, and show the armies to be very close. Running tests with searches to different depths or with different time
limits lead to wins by different sides. Even tiny tweeks of evaluation
parameters that seem like they should be unrelated can flip the victory
from one side to the other. I really have no idea which side is better,
although you are correct in being suspicious about unbalanced armies. I'm
sure if we could perform a perfect analysis we would discover that one side
is definitely better. But, it seems that they are so close that I'm pretty
confident that even slight miscalculations by the players will override any
inherrent unbalance.</p>
<p>Also, I should point out that using Zillions of Games as a tool for
computer analysis of chess variants is totally pointless. Zillions is an
impressive program, and is a very good tool for prototyping games and
getting a feel for how they play. Unfortunately, it is useless for
analysis because it is incapable of evaluating positions correctly. Even
in standard Chess it doesn't have a clue. It thinks that a Queen is worth
less than Rook+Bishop, when a thousand years of careful study have shown
that a Queen is worth more than the pair. And it can't even play a
reasonable opening; it loves to move the Queen to the center of the board
as early as possible. You can't really base anything on the results of
test-cases run by a program that makes really dumb moves right from the
start.</p>
<p>A new version of ChessV supporting Angels and Devils will be available
very soon, I hope. I'm still trying to stomp out one last annoying bug.
Of course, it is worth noting that any existing bugs could skew the
results of my computer analysis. Additionally, ChessV still relies on me
to provide evaluation parameters, and those could be wrong too. Sadly,
there is no perfect tool for testing such things. ChessV is still in its
infancy, but for the time being it's the best we've got.</p>