In the literature many researchers have tried to correlate the (dry) abrasion resistance of steels to their standard mechanical properties as determined in simple tensile and hardness testing. However, many unclear or even conflicting correlations have been reported, in part due to small data sets being used, different testing methods and relatively simple data processing techniques. In the present work 40 different low alloy steels were tested using the same ASTM G65 abrasion test and standard mechanical testing. The data set consists of 20 samples having the same chemical composition yet processed to different microstructures and different mechanical properties, as well as 20 steel grades with various compositions but having mechanical properties comparable to the first data set. The results show that for steels of the same composition significantly higher correlations between the abrasion resistance and certain mechanical properties (in particular strength coefficient K, hardness and UTS) are observed than those for steels having different compositions. In case both data sets were combined and a new regression analysis was applied, the quality of the correlation dropped significantly and the correlations became barely statistically relevant. Furthermore, when taking into account the measurement errors in both independent and dependent variables, the correlation coefficient became even worse and no statistically relevant correlation was observed at all. The current study clearly casts doubts on the validity of reported correlations between material (mechanical) properties and abrasion resistance.