<p>The aim of this study was to make performance comparisons of several Leeb hardness (LH) central tendency indicators with regard to their uniaxial compressive strength (UCS) prediction accuracies. Within the scope of the study, laboratory Leeb hardness tests (LHT) and UCS tests were performed on 24 drill core rock samples with different lithological origins and hardness ranges. The traditional arithmetic mean, median, 10% symmetric trimmed mean, 20% asymmetric trimmed median, interquartile range (IQR)-based median, and median absolute deviation (MADe)-based median were considered as the central tendency indicators. Their performance comparisons were made based on the normality and statistical distribution properties of the examined LH data sets. According to the results of the statistical analyzes, no significant differences among the derived UCS prediction models were found for normally distributed LH data (Shapiro-Wilk p-value &gt; 0.05 and/or skewness &lt; ± 1.0). However, for non-normality distributed LH data (Shapiro-Wilk p<i>-</i>value &lt; 0.05 and/or skewness &gt; ± 1.0), the central tendency indicators that are based on the interquartile range (IQR), median absolute deviation (MAD_e), and 20% asymmetric trimmed median outperformed the other central tendency parameters in terms of UCS prediction accuracy. These results suggest that, before adopting a particular central tendency indicator for LHT data evaluations, particular attention should be paid to the statistical distribution characteristics of the data set. In this context, the researchers and project engineers are encouraged to use robust test statistics for evaluating LH data due to their satisfactory performance under both normal and non-normal conditions.</p>