Abstract Thirteen sets of flexural strength data for simply supported beams were examined to check the appropriateness of the three-parameter Weibull distribution and the double exponential extremal distribution. As the thirteen data sets involve different kinds of ice with various microstructures, temperature, salinities and volumes tested under different loading rates, the parameters vary among the data sets. Kolomogorov-Smirnov one-sample goodness-of-fit test is found to be significant at 5% level of confidence for the two assumed distributions. The parameters as well as their standard deviations are given. It is found that the double exponential distribution is more convenient to work with as compared to the three-parameter Weibull form and it gives nearly equal mean strength with a more conservative estimate of the coefficient of variation.