Programming online judges are e-learning tools usually used in programming practices for the automatic evaluation of source code developed by students, for solving programming problems. Specifically, they contain a large collection of such problems where the students, at their own personalized pace, have to select and try to solve. Therefore, the increasing of the number of problems makes difficult the selection of the right problem to solve according to the previous users performance, causing information overload and a widespread discouragement. The current contribution proposes a recommendation approach to mitigate this issue by suggesting problems to solve in programming online judges, through the use of fuzzy tools which manage the uncertainty related to this scenario. The proposal evaluation, using real data obtained from a programming online judge, shows that the new approach improves previous recommendation strategies which do not consider uncertainty management in the programming online judge scenarios.