The reduced societal acceptance of living in regions exposed to snow avalanches, and the increased economic consequences when houses are located within a hazard zone, highlight the uncertainty concerning avalanche run-out prediction. The limitations of today’s zoning procedures are especially pronounced in potential avalanche terrain where there are few observations of snow avalanches, where old buildings are present in the potential run-out zone, and where the local climate does not favour severe snow accumulation. This paper combines a mechanical probabilistic model for avalanche release with a statistical/topographical model for avalanche run-out distance to obtain the unconditional probability of extreme run-out distance. For the mechanical model, a first-order reliability method (FORM) and Monte Carlo simulations are compared. The interpretation of the statistical/topographical model either as an extreme value model or as a single value model is discussed. Furthermore, both a classical approach where the probability of an avalanche occurring is a constant, and a Bayesian approach with stochastic probability, are compared. Finally, example applications in hazard zoning are presented, with emphasis on how the influence of historical observations, local climate, etc., on run-out distance can be quantified in statistical terms and how a specified certainty level can be found from constructing confidence intervals for, for example, the most likely largest run-out distance during various time intervals.