Serious games seem to be a promising alternative to traditional practices for learning. Recently, their use in computer science education and learning programming became more widespread. Actually, many students in programming courses have difficulties to master all required competencies and skills especially at introductory level and games have the potential to be an important teaching tool for their interactive, engaging and immersive activities. However, the critical point of a serious game remains the relation between the game and its pedagogical content. In fact, a serious game has to deal with two scenarios: the game scenario and the pedagogical scenario. In this paper, we introduce a system that aims to improve algorithmic learning using serious games. We propose a method for managing both game and pedagogical scenarios along with a double adaptation: the first phase that adapts the pedagogical scenario and the second phase that adapts the game scenario. We propose a software architecture that ensures the interaction between the aforementioned two phases. The learning and the game scenarios are planned according to the student's profile, his pedagogical path, his progress in the game and also taking into account the collected traces in the observation phase. [For the complete proceedings, see ED579335.]