Average deviation of experimentally determined phases from the final model phases. (a) The `SAD best centric' phase is the centric phase (0° or 180°) with the greater probability based on the SAD (single-wavelength anomalous dispersion) data. The `most centric SAD solution' is the phase (of the two ambiguous choices from SAD) that is closest to being centric (0° or 180°). The `SAD best' phase (the probability-weighted average phase) represents the standard choice for an ordinary (i.e. not quasi-racemic) SAD experiment. (b) A similar comparison restricted to reflections for which the model phase is within 20° of centric.