Focussing on extreme and skewed data, we investigate different methods, from non-parametric to Bayesian approaches, and identify assumptions under which some combination techniques are performing better. In this presentation, we apply multiple model combination methods to an ensemble of experts in a pooling approach and use the differences in outputs from the different combinations to illustrate how one can gain additional insight from using multiple methods. An illustration with extreme precipitation will be presented. Then, we present a generalized method, in order to fit skewed errors, and apply this new algorithm to a simulated insurance dataset.