Learning game strategy design through iterated Prisoner's Dilemma

The article investigates games strategies on the basis of experiments with iterated Prisoner's Dilemma, a classical non-zero sum game. The objective is to determine which strategies have the best chance of winning. Although some strategies, like tit-for-tat, emerge as better than others in some cases, it appears that there is no overall winning strategy, but that success or failure of individual strategies depends upon the strategies adopted by a population of opponents. Therefore, the winning strategy will change dynamically, and will need to be determined while the game in progress. Based on the results of this work, a strategy engine for games development is proposed, and a future development of strategy middleware is discussed