An Analysis of Hall-of-Fame Strategies in Competitive Coevolutionary Algorithms for Self-Learning in RTS Games Mariela Nogueira, Carlos Cotta, and Antonio J. Fern´andez-Leiva University of Computer Science, Havana, Cuba University of M´alaga, M´alaga, Spain mnogueira@uci.cu,{ccottap, afdez}@lcc.uma.es Abstract. This paper explores the use of the coevolutionary evaluation method Hall-of-Fame (HoF) in the application of competitive coevo- lution to find winning strategies in RobotWars, a two-player real time strategy (RTS) game developed at the University of Malaga. The main goal is to test different approaches in order to implement the concept of HoF as part of the self learning mechanism in competitive coevolutionary algorithms. Five approaches were designed and tested, they differ in the policies followed to keep the members in the champions’ memory during the updating process which deletes the weakest individuals, in order to consider only the robust members in the evaluation phase. It is demon- strated how strategies based on periodical updating of the solution set on the basis of quality and diversity provide globally better results. Keywords: coevolution, RTS game, game’s strategy, evaluation pro- cess, memory mechanism 1 Introduction Achieving bots that play in a similar way as a human user (i.e., not necessarily always winning, but exhibiting personality, creativity, and the ability to learn from mistakes) can be considered a variant of the famous Turing Test, and poses huge challenges for researchers in the field of Artificial Intelligence (AI). Evolutionary algorithms are a very appropriate approach in this matter, since they are capable of producing highly (possibly emergent) complex solutions as a result of the optimization process. Coevolution, as other biologically inspired techniques, is based on the interaction between different species, and represents one of the most interesting approaches to exploit in this area. Coevolutionary systems are usually based on two kinds of interactions: one in which different species interact symbiotically (i.e. the cooperative approach) and another in which species compete with each other (i.e. the competitive ap- proach). In cooperation-based approaches, an individual is typically decomposed in different components that evolve simultaneously and the fitness depends on the interaction between these components; in competition-based approaches, an