This study was carried out to determine genome sampling size of Cynodon accessions collected from Turkey for estimating genetic relatioships. One hundred and eighty-two Cynodon accessions collected in Turkey from an area south of the Taurus Mountains along the Mediterranean cost and ten known genotypes were genotyped using sequence related amplified polymorphism (SRAP) and the samples include diploids, triploids, tetraploids, pentaploids, and hexaploids. In this study, 182-markers were available from a previous work. We created 7 different data file, which include 5, 10, 20, 40, 60, 80, and 100 markers randomly selected from a 182-marker data file. First, similarity matrices was produced based on Dice' and, alternatively, simple matching' similarity coefficients for each data files. Then, a two-way mantel test was used to estimate Mantel correlation coefficients between each pair of similarity matrices. Mantel correlation coefficients between 5 and 10, 10 and 20,20 and 40, 40 and 60, 60 and 80, and 80 and 100 marker-based simple matching' similarity matrices were 0.718, 0.802, 0.826, 0.929, 0.977, and 0.975, respectively. Correlation estimates for the matrices based on Dice's procedure for the same data sets were 0.863, 0.916, 0.930, 0.966, 0.965, and 0.984. The lower number of markers (10 or 20) was required with Dice's procedure to achieve the higher correlation values (r > 0.9), and about random genome sampling size of 10 or 20 markers appeared to be sufficent in Cynodon accessions with a series of ploidy. This approach may apply to other plant species with ploidy series. (C) 2011 Elsevier Ltd. All rights reserved.