Fidel Castro has been a major Cuban politician of the XXth century; that influenced different left-wing regimes and political movements in Latin America, and around the World. His ability to seduce masses relied to a large extent, on his rhetorical abilities. Indeed, speech skills are crucial for a politician to captivate his audience to join his cause. Therefore, studying Castro's political speech, is a crucial step towards understanding his political success. Some studies addressed this task previously, mostly using small discursive samples, with only few speeches. However, using a small and possibly non-representative sample is likely to lead to biased results. To overcome this problem, the work presented here was carried out on a large corpus of 1,018 speeches and 7,548,480 words, combining state-of-the-art data mining tools, and the linguistic discourse analysis methodology of Patrick Charaudeau. Using the association of both techniques, we provide here a more representative characterization of Castro's main discursive strategies.