A comprehensive study on the stress-dilatancy behavior of cemented sand and its modeling is presented. The effect of confining pressure, relative density, and cement content on stress-dilatancy behavior are studied from the published experimental results and an additional series of experiments performed in this study. To facilitate a contrast and comparison of stress-dilatancy behavior between these datasets, a normalized stress ratio is proposed which removes the effect of mineralogy and morphology of parent sand. A set of key insights were obtained from this comparative study which aided in improving the stress-dilatancy relation; for example, the effect of initial conditions on stress-dilatancy behavior was found to be captured by the ratio of cohesion intercept (or tensile strength) and mean effective stress before shearing. The limitations of stress transformation, often used in modelling of cemented sand, were also systematically studied by a set of carefully designed experiments; it was found to be only applicable before gross yielding of cementation. After gross yielding, it is necessary to take in account of the breakage of bonds/cementation. The gross yield locus was identified from 70 experimental datasets and a cohesion/bond degradation model was formulated to model the stress-dilatancy behavior of cemented sand. The efficacy of stress-dilatancy relations (after including the gross yield locus and bond degradation behavior) is evaluated from the experimental results; the Rowe's stress-dilatancy relation was found to be most suitable with the proposed bond/cohesion degradation model.