This article presents a detail experimental procedure to perform float-and-sink tests for classifying coal samples according to their densities. Moreover, this article emphasizes obtaining ‘partition curves’ for three different coal samples (heavy media bath, big barrel and small barrel), which helps in evaluating and demonstrating classifier performance. Calculations of independent variables such as ‘Probable Error’ and ‘Imperfection’ are also discussed for partition curve that helps in evaluating the effectiveness of various beneficiation equipment used for the upgradation of quality of coal received from different coal mines. It was observed that there is a tendency for the partition curves to steepen as the density of separation decreases. In other words, separations at lower density is sharper than separations at higher density.