From kjahds@kjahds.com Fri Apr 19 11:32:25 1996
Date: Fri, 19 Apr 1996 11:32:03 -0400 (EDT)
From: Kenneth Albanowski <kjahds@kjahds.com>
To: perl-journal-contest@perl.com
Subject: Perl Prisoners Dilemma Contest Submission
Mime-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII



# Written by Kenneth Albanowski <kjahds@kjahds.com>
# Submitted by Kenneth Albanowski <kjahds@kjahds.com>

# Technique: Apply a gestalt technique to the Three-Way Prisoners Dilemma
#
# Explanation: The lowest (best) score possible in the TWPD is accessible
# only if you consistently testify against your friends, and they
# consistently hold out. This gives you a low score (possibly the lowest
# one), and them quite a large one.
#
# The key therefore is to a) guarantee that there are friends that will
# always hold out, and b) that other strategies that consistently testify
# won't be able to make use of your compatriots.
#
# A is solved via a gestalt technique: several contest entries working with
# each other to obtain scores that are not obtainable by a single entry.
# This has been accomplished by recruiting folks who decided did not enter
# their own entries in the contest, and asking them to submit this entry.
# (Note that collusion among authors was not ruled out.) All entries (as
# originally designed) should be identical, save for one entry that has the
# "we_are_master" flag set.
#
# If, when the gestalt strategy runs, the "we_are_master" flag is set, then
# it will consistently testify -- except for it's sixth response, where it
# will hold out. The master is not affected at all by the responses of it's
# friends.
# 
# If the "we_are_master" flag is not set, then this is a "slave" entry, and
# it will attempt to deduce whether it is competing against the "master"
# entry, to solve B -- this is accomplished by looking at the pattern of
# responsed from both friends. If either deviates from the known responses
# of the master, then it is known to not be the master. If neither friend is
# the master, then the slave will only respond with testify, thus increasing
# the non-gestalt competitors points as much as possible.
# 
# On the other hand, if the slave decides that either of it's friends is the
# master entry (or at least has not seen any information to the contrary),
# then it will always respond by holding out, thus producing the lowest score for
# the master.
#
# Note the important corollary: by lowering the master's score, the slave's
# scores are increased, indeed, it is not intended for any slave to win.
# Only the master can win. But it can only win _because_ the slaves are
# sacrified.
#
#
# The rules of this contest were unclear on who or what precisely wins. They
# say:
#
#     * The winning strategy will be the one with the lowest cumulative
#       score over all matches.
#
# This does not say whether the strategy needs to be a particular entry, or
# whether it can be the sort of meta-strategy that this gestalt embodies. I
# would like to argue that everybody who sent in this gestalt entry has won,
# not just the author, as such multiple entries are an implicit part of the
# strategy.
#
# - Ken

sub gestalt_ken {
    my ($my_choices_ref, $friend1_choices_ref, $friend2_choices_ref) = @_;
    my (@my_choices) = @$my_choices_ref;
    my (@friend1_choices) = @$friend1_choices_ref;
    my (@friend2_choices) = @$friend2_choices_ref;
	
	my($we_are_master) = 1;
	
	if($we_are_master) {
		return @my_choices==5?"H":"T";
	}
	
	if(@my_choices == 0) {
		return "H";
	}
	
	my($i,$fail1,$fail2);
	
	$fail1=$fail2=0;
	
	for($i=0;$i<@friend1_choices;$i++) {
		if($i==5 and $friend1_choices[$i] ne "H") {
			$fail1=1;
			last;
		}
		if($i!=5 and $friend1_choices[$i] ne "T") {
			$fail1=1;
			last;
		} 
	}
	if($fail1) {
		for($i=0;$i<@friend2_choices;$i++) {
			if($i==5 and $friend2_choices[$i] ne "H") {
				$fail2=1;
				last;
			}
			if($i!=5 and $friend2_choices[$i] ne "T") {
				$fail2=1;
				last;
			} 
		}
		if($fail2) {
			return "T";
		}
	}
	
	return "H";
}
__END__
--
Kenneth Albanowski (kjahds@kjahds.com, CIS: 70705,126)