Evaluating the efficiency of unguided search based on random walk in unstructured peer-to-peer networks is important because it provides guidelines in correctly setting the parameters of the search. Most existing work is based on simulations. We evaluate two analytical models – the algebraic model and the combinatorial model – for various search efficiency metrics against simulation results. We use the random graph topology and assume unguided searches. The results show that the two analytical models are accurate and match each other closely. We study the impact of the average node degree, hop count, number of walkers, and replication ratios on node coverage, object recall, and message efficiency, and on the accuracy of the models.