Studies attest that learning is facilitated if teaching strategies are in accordance with students learning styles, making learning process more effective and considerably improving students performances. In this context, one major research point--and a challenge--is to efficiently discover students' learning styles. But, the test and validation of new approaches in this field requires substantial amounts of financial, human resources (tutors and students) and time. In this way, the use of simulated students for test and validation of new approaches in this field is very important. Therefore, this work depicts the implementation and use of simulation for empirical evaluation of three different strategies for automatic learning styles modelling. It was necessary to compare the efficiency of the strategies, in order to choose the best one. Therefore, it was needed a practical mechanism to evaluate and compare them, preferably without the engagement of human resources. Then, the main goal of using simulation in this work is to compare strategies' efficiency and to discover the most promising one. The research was empirical, and has led to considerable enhancements on an intelligent component for automatic modelling of learning and teaching styles, which has been tested, adjusted and improved in a reasonable time and with low cost. The best strategy was clearly found, as depicted by experiments and results presented in this paper.