A probabilistic approach to algorithm analysis estimates the expected running time of a randomized algorithm, e.g. using a probability distribution of possible inputs.