Newsgroups: alt.gambling,rec.gambling.poker,rec.games.programmer,comp.ai.games,rec.gambling.misc,alt.sources,rec.games.design
Path: cantaloupe.srv.cs.cmu.edu!bb3.andrew.cmu.edu!newsfeed.pitt.edu!gatech!newsfeed.internetmci.com!howland.reston.ans.net!ix.netcom.com!netcom.com!mhall
From: mhall@netcom.com (Abdul Jalib M'hall)
Subject: Re: 5 cards stud poker logic/algorithm/strategy ?
Message-ID: <mhallDLyIn0.A4q@netcom.com>
Organization: NETCOM On-line Communication Services (408 261-4700 guest)
References: <cJp$mCEGv8$D078yn@hk.super.net> <Pine.PMDF.3.91.960122225915.541243530B-100000@minna.acc.iit.edu> <4eirk4$4np@tpd.dsccc.com>
Date: Mon, 29 Jan 1996 19:11:23 GMT
X-Original-Newsgroups: alt.gambling,rec.gambling.poker,rec.games.programmer,comp.ai.games,rec.gambling.misc,alt.sources,rec.games.design
Lines: 39
Sender: mhall@netcom3.netcom.com

In article <4eirk4$4np@tpd.dsccc.com>,
Darren Walker <dwalker@imtn.dsccc.com> wrote:
>Jim Karlinski (karljam@minna.acc.iit.edu) wrote:
>: Hah! If only there was such an algorithm! I'd be a rich man. The day a 
>: computer program can play poker competently is the day I chop my dick 
>: off. 
>
>Um...Should I mail you the butcher knife in preparation?
>
>All you need is a program that can calculate the basic statistics behind
>the game(and there are several of these...) and incorporate this into a 
>basic neural net.

As someone who made a living as an artificial intelligence researcher
and who is making income as a professional poker player, and who
has thought a great deal about implementing poker programs, I can
say that this simplification is laughable.  I'll dig up a quote 
from a game theory book when I get home.  The game is complicated
by the combinatorics of the cards and betting sequences (i.e., there
are a huge number of possible states), the incomplete information of
the opponents' hands, and the existence of N opponents with unknown
strategies (which may be collusional, unintentionally.)

I don't deny that a neural net might be useful, but you can't just
generate a handful of statistics, toss them into a neural net, and
expect to produce a good poker playing program.  Also, the neural
net, or any computer program, would almost by definition not supply
an optimal strategy in any sense, even for the much simpler heads up
case.  (The truth of this statement hinges on the number of unique
reachable states in a poker game, or more precisely the amount of
information necessary in the complete strategy.  For very simple poker
variants, optimal heads-up strategies are feasible, but for most real
poker games, representing the optimal strategy probably consumes more
storage space than is available on present day computers.)

-- 
Abdul  | Los Angeles is the place I gotta be, so I'm loadin' up the truck and
Jalib  | movin' to Beverly.  Hills, that is.  Hollywood Park, actually.
M'hall | Movie stars.  40-80 Hold 'em.
