Top Score in Axelrod Tournament

TL;DR

This paper develops and evolves finite state machine strategies to achieve the highest scores in the Axelrod Tournament, demonstrating significant improvements through evolutionary algorithms.

Contribution

It introduces an evolving FSM-based approach that outperforms existing strategies in the Axelrod Tournament, achieving top rankings.

Findings

Final FSM strategy ranks first in the Axelrod Tournament.

Evolutionary algorithms effectively optimize FSM strategies.

Strategy outperforms top existing strategies in both short and long run.

Abstract

The focus of the project will be an examination of obtaining the highest score in the Axelrod Tournament. The initial design of the highest score in the Axelrod Tournament consisted of looking at the Cooperation rates of the top strategies currently in the Axelrod Library. After creating an initial Finite State Machine strategy that utilized the Cooperation Rates of the top players; our ten-state FSM finished within the top 35 players out of 220 in the short run time strategies in the Axelrod Library. After a quick evolutionary algorithm, 50 generations, our original ten-state FSM was changed into an eight-state FSM, which finished within the top 5 of the short run time strategies in the Axelrod Library. This eight-state FSM was then evolved again using a full evolutionary algortihm process, which lasted 500 generations, where the eight-state FSM evolved into a another eight-state FSM which finishes first in the Axelrod Library among the short run time strategies and the full Axelrod Tournament. From that final FSM, two of the eight states are inaccessible, so the final FSM strategy is a six-state FSM that finishes first in the Full Axelrod Tournament, against the short run time strategies as well as the long run time strategies.