Many methods exist for estimating species composition, but few studies compare those useful in improved horse pastures. The objective of this study was to examine 4 techniques for estimating desirable forage species composition in 2 cool-season horse pastures based on prevalence estimates, repeatability, bias, and practicality, and to select a method for use in a subsequent grazing study. The techniques included Equine Pasture Evaluation Disc (EPED), Line-Point Intercept with 3 transects of 50 observations each (LPI 3-50), LPI with 5 transects of 30 observations each (LPI 5-30), and Step Point (StPt). A generalized linear-mixed effects model procedure of SAS (GLIMMIX) with a logit link was used to test for differences among each species separately. When methods were significantly different (α = 0.05), pairwise comparisons were performed using a paired t-test. The methods did not differ in detecting creeping bentgrass (P = 0.3334) or orchardgrass (P = 0.4207), but there were differences for Kentucky bluegrass (P = 0.0082), tall fescue (P = 0.0314), and other (P = 0.0448). Repeatability plots displayed lower method repeatability as species prevalence increased. Agreement was analyzed between pairs of methods by grass species. Five out of 30 pairs showed significant overall bias (P = 0.0114, 0.0045, 0.0170, 0.0328, and 0.0404), and 3 of them were between LPI 3-50 and EPED. The LPI 3-50 and LPI 5-30 techniques agreed perfectly in prevalence and bias, as did StPt and EPED, meaning they can be used interchangeably. The techniques LPI 3-50 and EPED were the most dissimilar methods. In conclusion, StPt can be used interchangeably with LPI, but StPt was selected due to its thorough representation of the pastures and ease of use.