Few studies have been conducted for susceptibility of rock falls in mountainous areas. In this study, we compare and evaluate rock fall susceptibility mapping using bivariate statistical [weight of evidence (WoE)], analytical hierarchy process (AHP) and frequency ratio (FR) methods along 11 km of a mountainous road in the Salavat Abad saddle in southwestern Kurdistan, Iran. A total of 34 rock fall locations were constructed from various sources. These rock fall locations were then partitioned into a training dataset (70% of the rock fall locations) and a testing dataset (30% of the rock fall locations). Eight conditioning factors affecting on the rock falls including slope angle, aspect, curvature, elevation, distance to road, distance to fault, lithology and land use were identified. The modeling process and rock fall susceptibility mapping has been constructed using three methods. The performance of rock fall susceptibility mapping was evaluated using the area under the curve of success rate curve for training and prediction rate curves (PRC) for testing datasets and also seed cell area index. The results show that the rock fall susceptibility mapping using the WOE method has better prediction accuracy than the AHP and FR methods. Ultimately, the weight-of-evidence method is a promising technique so that it is proposed to manage and mitigate the damages of rock falls in the prone areas.