The restrained ring test is frequently used to assess the susceptibility of a concrete mixture to restrained shrinkage cracking. The test method has been recently standardized as ASTM* C 1581 (2004). Despite many useful applications of this test, concerns have been raised with respect to the repeatability and interpretation of the results when some of the rings made from a concrete mixture crack, while other specimens from the same mixture do not exhibit cracking at all. This paper provides explanation on why even in a properly performed experiment, not all the rings are always expected to crack. The work presented in this paper combines experimental and stochastic approach to analyze the repeatability of the test method and quantify variability in shrinkage measurements. In each of the experiments performed in this work, six rings were cast simultaneously from each batch of mortar prepared. Two water-tocement ratios were considered: w/c=0.30 and w/c=0.40. Additionally, the effect of shrinkage reducing admixtures on variability in the time of cracking was evaluated. The results of the experiment and the simulations are presented to explain why in a given experiment not all the rings may necessarily crack. Additionally, a probabilistic approach is used to describe the probability of cracking in a restrained concrete element. The effect of variability in material properties is included and its effect on concrete cracking prediction explained. This work can be further used by engineers as a tool to evaluate different materials performance and deliver cracking prediction. Once information about cracking probability is obtained, material properties or mixing procedures can be modified and potential for cracking minimized. As such, this approach meets the need for accurate assessment tools that can be implemented in performance-related specifications.