This paper presents a probabilistic approach in sizing distribution transformers capacity supplying residential, commercial, and combination of residential and commercial loads. Load data recorded from the three categories of aggregate loads are used to establish probability distribution curves based on the goodness-of-fit test. The gamma distribution is found to be the best fit among all probability distributions for kilo-volt-ampere (KVA) recorded at peak periods. It is shown that distribution transformers capacity could be more optimally sized by considering a risk factor and thus results in cost savings.