Newsgroups: comp.ai.genetic,comp.ai.alife
Path: cantaloupe.srv.cs.cmu.edu!rochester!udel!news.mathworks.com!news.ultranet.com!news.sprintlink.net!cs.utexas.edu!oakhill!particle!nahas
From: nahas@particle.sps.mot.com (Michael Nahas)
Subject: Re: Prisoner's Dilemma
X-Nntp-Posting-Host: particle.sps.mot.com
Message-ID: <DAsACx.39z@oakhill.sps.mot.com>
Sender: news@oakhill.sps.mot.com (oakhill news)
Organization: SSDT SPS, Motorola Inc., Austin TX
References: <BrianS-2306950136200001@slip-1-92.ots.utexas.edu> <Ted.Belding-2406951449220001@pm056-27.dialip.mich.net> <Pine.ULT.3.91.950625021800.7629A-100000@rac8.wam.umd.edu>
Date: Mon, 26 Jun 1995 14:17:20 GMT
Lines: 26
Xref: glinda.oz.cs.cmu.edu comp.ai.genetic:6202 comp.ai.alife:3703

Tit-for-tat can be used for Repeated Prisoner's Dilemma even if
the number of the matches is known.  For shorter runs of matches,
it does reduce to alway Defect, but for longer runs it is possible
to get a long run (or runs) of Not Defect which eventually result
in Defect near the end of the contest.  The reason is the payoff
for a long run of Not Defect on both sides is much larger than a
long run of Defect on both sides.  Once that value is large enough,
it is worth risking some small piece if a long run of Not Defect
can be achieved.

The reason some contests run for an undefined number of matches is
to remove the small advantage from being the first to Defect at the
near the end of the contest.  

I believe Steve Levy's book describes both contest that were run 
by humans (not computers) which resulted in Tit-for-tat winning.  
The first was of constant lenght, which in your opinion should have 
been won by a contestant that constantly defected, but was won
by Tit-for-tat.  

If you have any questions about this, please email me at
nahas@ssdt-oakhill.sps.mot.com

Michael Nahas


