Abstract The estimation of the coancestry coefficient associated with a pair of individuals, given their genetic types, has been a matter of great theoretical and practical importance, namely in forensic kinship tests. Generally considered as a condensed measure of relatedness, it is quite often assumed that there is a loss of information when this measure is taken in consideration instead of Jacquard's identity-by-descent partitions. Nevertheless, considering non-inbred individuals, it can be proved that, excluding parent–child and full-siblings pedigrees, both measures are equivalent for genealogies that relate two individuals by exactly one path. In this work, the coancestry coefficient between two (not necessarily non-inbred) individuals is straightforwardly inferred from their genetic types through the probability of finding two identical alleles, each one randomly chosen from each individual, or, alternatively, through the expected homozygosity of their virtual offspring. The presented method is rooted in the algebraic expression for the probability of finding a homozygous individual in the population. It is shown that the coancestry coefficient of two individuals related by a given kinship can be mathematically modeled by a linear function depending on (a) the expected homozygosity of their virtual offspring, and (b) the average homozygosity in the population; the slope and the y -intercept of the straight line carrying information about the gene diversity of each database.