When DNA profile comparisons between a crime scene trace and a reference sample generate correspondence, the match probability has to be estimated, so that evaluation of the strength of the forensic DNA evidence can be made. The random match probability estimations require information on allele frequencies and an adjustment factor, referred to as theta (θ) or Fst, a co-ancestry correction factor for subpopulation effects. The θ value has been standardized for urban and isolated populations, but inconsistencies have been reported when it is specifically calculated for smaller and isolated populations, including Amerindian populations. Notably, attempts to characterize forensic markers of these minor populations have been extensively limited and more conservative estimates of the correction factor may be generated for each of them. Therefore, we estimate allele frequencies of 21 autosomal STR markers used for forensic testing and calculated relevant forensic parameters for the set. In addition, we featured the possible structure of five Brazilian Amerindian populations that have been genetically isolated for centuries so we could obtain the appropriate θ value for them. The sample consisted of 319 individuals: (1) 121 Kaingang, from two communities: Ivaí (KIV=61) and Rio das Cobras (KRC=60); and (2) 198 Guaranis from three communities: Mbya from Rio das Cobras (GRC=51), Guarani Ñandeva (GND=71) and Guarani Kaiowá (GKW=76). Between Guarani populations low (Rst=0.0402, p < 10−4) to high (Rst=0.1557, p < 10−5) differentiation was found. Regarding Guarani and Kaingang populations, intermediate (Rst=0.0590, p < 10−5) to high (Rst=0.1604, p < 10−5) differentiation was found. The two Kaingang populations showed very low differentiation between them (Rst=0.0017, p = 0.27), which justifies the union of both genetic data for forensic databases and calculations. The combined power of discrimination (PD) and the combined power of exclusion (PE) were calculated for each population, demonstrating the usefulness of this set of markers in forensic and kinship analysis regarding these populations. Considering the demographic heterogeneity of Amerindian populations in general, the Fst mean value (0.03) was evaluated regarding 43 different indigenous populations from the Americas, including Guaranis and Kaingangs. This result confirms the adequacy of the standardized θ value for the forensic random match probability estimations involving Amerindian populations.