Repeated Prisoner's Dilemma: We don't need another hero...
Hoisted from Comments:
Grasping Reality with Both Hands: Brad DeLong's Semi-Daily Journal: A Note on a GRIM Game of Repeated Prisoner's Dilemma...: "But why restrict ourselves to Axelrod's 63 strategies? Why not follow Linster (1990, 1992) in beginning with all finite automata with at most two states?"
Because "all finite automata with at most two states" is just a fancy and round-about way of saying "all strategies with only a single turn memory". GRIM beats TFT in this environment because no strategy has a long enough memory to distinguish punishment from exploitation.
It's also worth noting that GRIM doesn't have a "Road Warrior" effect. Since GRIM is a "nice" strategy, a mostly GRIM population is characterised by mutual cooperation.
Posted by: pete | February 08, 2007 at 03:54 AM
GRIM is "nice" to itself and to other don't-defect-first strategies that it cannot distinguish from itself. GRIM wages total thermonuclear war against everything as soon as it recognizes that what it is playing is not-GRIM. That seems Mad Max-like to me...