In this article the issues are discussed with the Bayesian approach, least-square fits, and most-likely fits. Trying to counter these issues, a method, based on weighted confidence, is proposed for estimating probabilities and other observables. This method sums over different model parameter combinations but does not require the need for making assumptions on priors or underlying probability functions. Moreover, by construction the results are invariant under reparametrization of the model parameters. In one case the result appears similar as in Bayesian statistics but in general there is no agreement. The binomial distribution is also studied which turns out to be useful for making predictions on production processes without the need to make further assumptions. In the last part, the case of a simple linear fit (a multi-variate example) is studied using the standard approaches and the confidence weighted approach.