The aim of this paper is to study why we should have a closer look to Pittsburgh-style Classifier-Systems (Pitt-CS). This kind of classifier-systems were introduced by Smith during the early 80’s and was nearly forgotten during 20 years. We revisit those kind of classifiers adapting them. We choose as background of our study the ‘El Farol’ bar problem introduced by Arthur. This multi-agents system problem leads us to test several abilities as memory, problem response and uniformity of population, using GA-independent parameters. Results have shown that adapted Pitt-CS have useful abilities.