The two systems currently used in Australia for assessing the body reserves of live sheep (fat and condition scoring) were evaluated for their ability to quantify the range of scores within a mob and their ability to discriminate between sheep of varying fatness. Three ewe genotypes (fine Merino, medium Merino and first-cross Border Leicester × Merino) were independently assessed by four assessors experienced in each system (i.e. eight assessors). Twenty-four hours after assessment the ewes were slaughtered in a commercial abattoir and hot carcass measures of tissue depth at the GR site (thickness of tissue over the 12th rib, 110 mm from the midline) taken to allow correlation with the assessed live scores. There was at least three times more phenotypic variation in fat score assessment compared with condition score as the fat score assessors used a wider range of scores. Despite this the average repeatability of fat scoring (0.73–0.85) was greater than condition scoring (0.64–0.84) particularly within the fine Merino genotype. Although there was a strong linear relationship between assessments of fat and condition scoring, the regression coefficients comparing the two scoring systems indicated a greater discrimination among animals on fat score across all three genotypes. Furthermore the fat scores had a significantly higher correlation with GR tissue depth measured on the carcass than condition scores. These data indicate that both systems will similarly assess the average body fat reserves of a mob or animals; however, fat scoring achieves greater discrimination when the goal is to identify individuals that are lower or higher than the mob average.