Comments/Ratings for a Single Item

~~Earlier~~ ⇧Reverse Order⇩ ~~Later~~

Roberto Lavieri wrote on Mon, Nov 15, 2004 12:58 PM EST:

Mr. Joost Aan de Brugh has gently implemented this new ZRF which enforces
the Checkmate condition, and in fact, all winning threatenings are defined
as Check. All cases are considered in a logical manner. The old ZRF
generates some artificial moves with the object of calibrate the pieces
values, in the new ZRF this is not performed, and pieces are valued
incorrectly, by example, Withdrawer values is up to five times the
Coordinator value, and three times the Chameleon value. In the same way,
other assigned values by Zillions are not correct. Some corrections are
still needed to obtain a ZRF that can play this game a little better,
although Zillions is not a very strong player anyway, I can say the old
ZRF played Maxima at a level around 1800, if there is possible a
translation of FIDE-ELO here. The new ZRF 'ELO' is, subjectively, around
1600.

Joost Brugh wrote on Sun, Jan 22, 2006 06:58 PM EST:

I see that my ZRF significantly undervalues Coordinators. You can fool Zillions and make it think that a piece is more valuable by adding useless add's. The problem with the old ZRF is that the Coordinator is not obliged to capture the coordinated enemy pieces. If you do:

... (capture ThisAndThat) add add add add ...

ThisAndThat is only captured in the first add. This action is canceled after that add. The three other add's do a move without performing the capture-action. Maybe it also works if you even further fool Zillions by adding 'completely useless add's'

(piece (name Coordinator) ... (moves (move-type Regular) (CoordinatorMove n) ... (move-type Monkey) (add add add add add add) ) )

And say in the turn-order that White and Black do only Regular moves. I don't know much about Zillions AI. But maybe this works.

Antoine Fourrière wrote on Sun, Jan 22, 2006 08:41 PM EST:

I think both tricks are useful.

Move priorities ensure that the AI needs not look at all for the extra
moves.
But the bogus square (mine are defined with a position of (-1 -1 -1 -1))
and the verify diminish the value of each add and thus allow more
precision in the points you intend to give.

(For my recent Dual Chess, which uses them both, I also defined several
pieces twice, so that the AI values them according to position, while
identical graphics and the use of the translate command make them appear
similar.)

3 comments displayed

~~Earlier~~ ⇧Reverse Order⇩ ~~Later~~

Permalink to the exact comments currently displayed.