The Arc Routing Problem is a kind of NP-hard routing problems where the demand is located in some of the arcs connecting nodes and should be completely served fulfilling certain constraints. This paper presents a hybrid algorithm which combines a classical heuristic with biased random sampling, to solve the Capacitated Arc Routing Problem (CARP). This new algorithm is compared with the classical Path scanning heuristic, reaching results which outperform it. As discussed in the paper, the methodology presented is flexible, can be easily parallelised and it does not require any complex fine-tuning process. Some preliminary tests show the potential of the proposed approach as well as its limitations