Assigning a confidence measure is a challenging stochastic inference problem. Some algorithms only yield the predicted value without evaluating the measure of confidence over the decision. Support vector machines is one algorithm that showed state of the art decision accuracy but lacks a measure of confidence over the decisions. In this paper we propose a confidence measure based on the VC (Vapnik and Chervonenkis) dimension of a learning algorithm. The resulting confidence measure is then tested on the well known US postal handwritten digit recognition. The results show high and improved correlation between the decision and the confidence measure.