A.I. applied to card games

G

Guest

Guest
Archived from groups: comp.ai.games (More info?)

I'm looking for documentation regarding A.I. applied to card games such as
bridge,hearts and, in particular, whist.
Can anyone suggest me some good site on the web?

These are the rules of Whist:
52 cards deck
4 players

in the first round the mace-bearer gives 1 card to each player and turn
another one as a trump.
in the second round the number of cards are 2 each, the third round 3 and so
on till the 12th round with 12 cards each.
then 4 rounds with 13 cards each are played. In this rounds there is no
trump.
then you start going back: one round with 12 cards, one with 11... till the
last round with 1 card.
Before the round begins,each player, starting from the one next to the
mace-bearer, has to declare how many hands he thinks he'll win.
if the player takes exactly the hands he declared, he scores 10 points plus
the hands he declared, otherwise he gets only 1 point for each hand he took.
Ace is the highest card.
the player who begins can play every card he wants, and the other players
have to play a card of the same suit, if they have one. Otherwise they can
play another card (likely a trump).
The highest card on the table wins, unless there is one or more trumps: in
that case the highest trump wins.
Who wins the hand starts to play the next one.
That's all.
Obviously, who has the highest score at the end of the 28 rounds wins the
game.



I play this game with my friends but I'm trying to create a software for
playing against the computer.
Any help is welcome :)

Alessandro Bognetti
Italy
 
G

Guest

Guest
Archived from groups: comp.ai.games (More info?)

"Alessandro" <alebog@libero.it> wrote in message
news:VHSfd.4261$Es2.88836@twister2.libero.it...
>

> I play this game with my friends but I'm trying to create a software for
> playing against the computer.
> Any help is welcome :)

A suggestion for an approach to "What card should I play?"

What would be the ideal card to play? Rank them top to bottom.
At this point, one could select from the highest ranked card in one's hand.

There's a couple other matters though; one is memory. We have the additional
information of what cards where played by whom, so we know ahead of time if some
cards are not actually beaten, and which ones are still possible. This could be
used to adjust the ranking. The other matter is that we might not want to play
an "idealistic" determinist strategy. Would you get bored playing against
predictable opponants?

So rather than selecting the ideal card, we could view that choice as the "most
probable" card, and then have rapidly diminishing probability of choosing the
next-best choice, etc. This may be a good tactic in the game as well, but you
would know best on that.

-:|:-
AngleWyrm
 
G

Guest

Guest
Archived from groups: comp.ai.games (More info?)

Alessandro wrote,

> These are the rules of Whist:

That's not whist, Alessandro. Whist is quite a different game; bridge
was derived from it.

What you describe is something like Oh Hell. Oh Hell is a great game.
Everyone I ever taught it to enjoyed it. What you described doesn't
look like as much fun as what I used to play as Oh Hell. The rules I
used are as stated at the URL shown below, except that I didn't treat
scoring a bid of zero any differently from any other bid (so I scored
a successful bid of zero as ten points).

The URL is http://lynx.dac.neu.edu/n/ncheung/cardgames/ohhell.html

Oh Hell would be an interesting game to write AI for, except that you
should have at least three players.

--
Dan Amodeo
E-mail: take my last name, all lower case, put a seven in the middle,
then add at earthlink dot net
 
G

Guest

Guest