Inspired by the statistical/topographical Norwegian approaches and ideas of Ancey and Meunier, this paper presents a method to compute the predictive distribution of snow avalanches runout distances and uncertainties associated with design values. A very simple propagation operator and minimal statistical hypothesis are involved. Accordingly to the French historical database, only release and runout altitudes are necessary. Main contribution of the work is the proposed statistical model that is flexible enough to take reasonably avalanche data variability in account and to express inter variables correlations. The bayesian framework facilitates parameter estimation and allows taking estimation error in account for predictive simulations.