Accuracy and Precision

Accuracy
The accuracy of a computational approach determines the closeness of a calculated value to its true (ideally: experimentally measured) value. In order to make the assessment of the accuracy of a particular theoretical method independent of a particular data point, it is common practice to define a larger data set for benchmarking purposes. A well known example for this strategy is the G3/05 thermochemical data set comprising 454 data points (L. A. Curtiss, P. C. Redfern, K. Raghavachari, J. Chem. Phys., 2005, 123, 124107).

Precision
The precision of a computational approach determines the reproducibility of a calculated value by repeated calculation. The precision of a particular computational approach may suffer from inadequate implementation of algorithms or improper selection of user-defined parameters. This latter category involves issues such as the number of SCF cycles, the accuracy of integral evaluation, the selection of integration grids in grid-based DFT calculations, the convergence criteria for geometry optimizations etc.