Full FEM models with incremental crack propagation usually require very small crack propagation increment lengths Δa to accurately predict crack paths. Beside the criterion for the crack propagation direction, the approach to evaluate this propagation direction is important: the driving force of the current crack can either be used directly (explicit approach) or iteratively adapted test cracks can be introduced (implicit approach). Furthermore, the shape of the crack increments can be either straight or curved. Considering two 2D plates with circular holes, the accuracy of the predicted crack paths as a function of Δa is here evaluated for an explicit approach with straight crack increments (explicit approach), an implicit approach with straight crack increments (straight approach), and an implicit approach with curved crack increments (curved approach). It is shown, that the explicit approach and the straight approach react too late and too early, respectively, to spatial changes in the stress field. This work proposes a means of dealing with the changing crack driving force direction within the crack propagation increments for the explicit and straight approaches. All approaches are benchmarked by comparing their efficiency, which is evaluated as the mean deviation from a reference crack path per number of FEM computations required. It is shown that the use of curved crack increments is very efficient in some cases, but less so in one of the cases considered. A significant improvement in efficiency is shown by our adaptation of the explicit and straight approaches.