Check out Glinski's Hexagonal Chess, our featured variant for May, 2024.


[ Help | Earliest Comments | Latest Comments ]
[ List All Subjects of Discussion | Create New Subject of Discussion ]
[ List Earliest Comments Only For Pages | Games | Rated Pages | Rated Games | Subjects of Discussion ]

Single Comment

[Subject Thread] [Add Response]
H. G. Muller wrote on Mon, Apr 20, 2009 03:08 PM UTC:
Mats: 'But it's to few games. However, it could be a program problem.
Maybe Zillions is intelligent enough to handle this piece well.'

Well, 3.5-0.5 is indeed far to few games to conclude anything. It could
easily occur in a match between two exactly equal opponents if one of them
is slightly lucky (e.g. because his opponent blunders away a single win).

I am at 366 games now, and the Guanacas lead by 57% over the Alpacas. When
I reach 400 games there, I will delete the f-pawn of the Guanaca side, and
see how they do then. (The Alpacas should win then by a similar amount.)
Two Knights vs two Guanacos is at about 250 games now, and the Knights lead
by 77%.

Intelligence is usually not in the vocabulary of Chess programs. It is all
brute search power, going through millions of positions with a very
simplistic evaluation. (Counting wood or moves.) Especially for a
generalist program like Zillions, that has no specific guidelines
programmed in for the Guanaco for sure.

I don't believe that handling Knights (or Alpacas, for that matter)
requires less intelligence than handling Guanacas. I don't believe that
Zillions would be any better at handling any specific piece than Fairy-Max.
Fairy-Max is only a very simple Chess program, but despite its complete
lack of programmed knowledge (except the piece values) it plays
surprisingly strong in normal Chess, dominating over many engines that are
stuffed with Chess-specific knowledge. Apparently it does not need any
knowledge to handle its pieces intelligently enough to win; plain search is
good enough, even in the end-game.

From what people told me, I don't think Zillions would be a match for
Fairy-Max in Chess variants with FIDE-like Pawns that they can both play. I
have not tested this myself, of course, as Zillions is commercial software
that I don't have. So it is just based on what people that do have
Zillions told me.

I understand that there is an adapter that allows WinBoard engines to play
in the Zillions GUI. Isn't it possible to play Fairy-Max against Zillions
automatically, that way? Then we could handicap one of them by time odds
until they play equally strong in normall Chess, and then add or substitute
Guanacas on both sides, to see if this breaks the equivalence because one
side handles the Guanacas 'more intelligently' than the other.