The authors define the requirements and a conceptual model for comparative evaluation research of simulation games and serious games (SGs) in a learning context. A first operationalisation of the model was used to comparatively evaluate a suite of 14 SGs on varying topics played between 2004 and 2009 in 13 institutes of higher education in the Netherlands. The questions in this research were: what is the perceived learning effectiveness of the games and what factors explain it? How can we comparatively evaluate games for learning? Data were gathered through pre- and post-game questionnaires among 1000 students, leading to 500 useful datasets and 230 complete datasets for analysis (factor analysis, scaling, "t"-test and correlation analysis) to give an explorative, structural model. The findings are discussed and a number of propositions for further research are formulated. The conclusion of the analysis is that the students' motivation and attitudes towards game-based learning before the game, their actual enjoyment, their efforts during the game and the quality of the facilitator/teacher are most strongly correlated with their learning satisfaction. The degree to which the experiences during the game were translated back into the underlying theories significantly determines the students' learning satisfaction. The quality of the virtual game environment did not matter so much. The authors reflect upon the general methodology used and offer suggestions for further research and development. (Contains 14 tables, 3 figures, and 4 notes.)