[net.math] Prisoner's Dilemma

colonel@gloria.UUCP (George Sicherman) (05/09/84)

[YOU are number SIX.]

I think that _Scientific American_ eventually held a contest, where
all sorts of people submitted Prisoner's Dilemma programs - some of
them very complex.  The winner was Rapoport, with a program that
merely

	1. cooperated,
	2. did what the opponent just did,
	3. did what the opponent just did,
	...

Can anybody confirm my recollection?
-- 
Col. G. L. Sicherman
...seismo!rochester!rocksvax!sunybcs!gloria!colonel

bill@utastro.UUCP (05/10/84)

>  I think that _Scientific American_ eventually held a contest, where
> all sorts of people submitted Prisoner's Dilemma programs - some of
> them very complex.  The winner was Rapoport, with a program that
> merely
> 
> 	 1. cooperated,
>  	 2. did what the opponent just did,
> 	 3. did what the opponent just did,
> 	 ...
> 
> Can anybody confirm my recollection?

Yep.  He called the strategy "tit for tat".
-- 

	Bill Jefferys  8-%
	Astronomy Dept, University of Texas, Austin TX 78712   (USnail)
	{ihnp4,kpno,ctvax}!ut-sally!utastro!bill   (uucp)
	utastro!bill@ut-ngp			   (ARPANET)

gwyn@brl-vgr.ARPA (Doug Gwyn ) (05/11/84)

(Re: ScAm contest of PD playing programs)

Ah, but the true Prisoner's Dilemma game is played just once.  This is an
important technical point that can indeed affect one's strategy.

Note that the "cooperate at first, then do what the other player did
last round" strategy loses relative to the "use dominant pure strategy"
approach whenever there IS a dominant pure strategy (actually it might
tie if it gets lucky, but it cannot beat the pure strategy).

	I'm not a number, I'm a free man!
		Ha, Ha, Ha ...