Correctly inferring relatedness among samples is essential for genetic analysis. It can be helpful for preventing false signals in genetic association studies and finding relatives in forensic genetics. However, relatedness among samples is not always obtained when collecting sample data; in most cases, the relatedness is unknown and needs to be determined. Here, we develop an algorithm to infer relatedness among samples that we aim to be more efficient than current related approaches such as PLINK. Our approach is based on finding stretches of shared alleles across windows of moderate length (3-5 centiMorgans) in the genome. With this information, the algorithm infers the degree of relatedness up to the third degree from the number of windows that are similar between a pair of individuals.