Joints are the main features encountered in rock and sliding of rock blocks on joints is classified as the principal source of instability in underground excavations. In this regard, joints’ peak shear strength is the controlling parameter. However, given the difficulty in estimating it, shear tests are often performed. These are often quite expensive and also time consuming and, therefore, it would be valuable if shear tests could be artificially performed using numerical models. The objective of this study is to prove the possibility to perform virtual numerical shear tests in a PCF2D environment that resemble the laboratory ones. A numerical model of a granite rock joint has been created by means of a calibration process. Both the intact rock microparameters and the smooth joint scale have been calibrated against macroparameters derived from shear tests performed in laboratory. A new parameter, the length ratio, is introduced which takes into account the effective length of the smooth joint compared to the theoretical one. The normal and shear stiffnesses, the cohesion and the tensile force ought to be scaled against the length ratio. Four simple regular joint profiles have been tested in the PFC2D environment. The analysis shows good results both from a qualitative and from a quantitative point of view. The difference in peak shear strength with respect to the one computed with Patton´s formula is in the order of 1% which indicates a good accuracy of the model. In addition, four profiles of one real rough mated joint have been tested. From the scanned surface data, a two-dimensional profile has been extracted with four different resolutions. In this case, however, interlocking of particles along the smooth joint occurs, giving rise to an unrealistic distribution of normal and shear forces. A possible explanation to the problem is discussed based on recent developments in the study of numerical shear tests with PFC2D.