GIAO NMR chemical shifts have been calculated for a set of 28 pairs of diastereoisomers in order to test the ability of NMR shift calculation to distinguish between diastereomeric structures. We compare the performance of several different parameters for quantifying the agreement between calculated and experimental shifts from the point of view of assigning structures and introduce a new parameter, CP3, based on comparing differences in calculated shift with differences in experimental shift, which is significantly more successful at making correct structure assignments with high confidence than are the currently used parameters of the mean absolute error and the correlation coefficient. Using our new parameter in conjunction with Bayes’ theorem, stereostructure assignments can be made with quantifiable confidence using shifts obtained in single point calculations on molecular mechanics geometries without computationally expensive ab initio geometry optimization.