It is well known for contractors that due to volume change during the hydration phase in concrete structures, large stresses can arise if the structures are restrained, stresses that may cause extensive cracking. Crack risk estimation of early age concrete structures can be based on five steps. Firstly, the type of structure, the material proportions and possible measures to avoid cracking have to be chosen. Secondly, the temperature development has to be determined, and thirdly the restraint situation. Fourthly, structural calculation of the stress or strain ratios follows, which in the fifth step are compared to stated partial coefficients that should not be exceeded. The restraint situation is one of the crucial factors in the crack risk analyses. A semi-analytical method has been derived for the determination of the restraint variation in early age concrete structures and especially for the case wall on slab. The method is derived using compensated line theory. The model depends on the geometry of the structure, the boundary restraint situation, and the location of the young parts on the old parts. The model is supplemented with the effects of high walls and the effects of short structures and/or possible slip failure in the ends of the joint between the young and the old concrete. The model is by regression technique compared to almost 3000 3D elastic FEM calculations of the restraint variation in walls on slabs with different dimensions and base restraint situations. The effective width of the slab is introduced as the only adjustment parameter to get the model to correspond with the FEM calculations. Partial coefficients for thermal cracking problems of young concrete have been calculated and compared with the values stated in the Swedish building code for bridges. The code values are only based on experiences and logical reasoning, whereas the calculated values form a more theoretical base for their determi-nation. The coefficients are calculated with a probabilistic method. Various possible variations of the used variables have been studied showing the wide range of possible results depending on the input. However, with use of material properties and reasonable assumptions related to thermal cracking problems, fairly good agreement has been found between the stated values in the Swedish code and the values obtained through the probabilistic method.