This article compares two edge-based Estimation of Distribution Algorithms named Edge Histogram Based Sampling Algorithm (EHBSA) and Coincidence Algorithm (COIN) in multimodal combinatorial puzzles benchmarks. Both EHBSA and COIN make use of joint probability matrix of adjacent events (edge) derived from the population of candidate solutions. These algorithms are expected to be competitive in solving problems where relative relation between two nodes is significant. The experiment results imply that EHBSAs are better in convergence to a single optima point, while COINs are better in maintaining the diversity among the population and are better in preventing the premature convergence.