Determination of the heavy-atom substructure shift required to bring the pseudo center of inversion of the crystal to the origin. The correct shift is based on the expectation that the protein phases should be approximately centric for a quasi-racemic crystal. The two-dimensional contour plot shows the average deviation of the more nearly centric phase (among the two ambiguous choices for each reflection based on SAD data) from 0° or 180°. Each point describes a candidate coordinate shift of the heavy-atom substructure origin in the xy plane. (a) For the snow flea antifreeze protein (Pentelute et al., 2008), the structure was reported with the pseudo center of inversion at the origin. The correct shift for the anomalous substructure from its reported position is therefore (0, 0, 0). The existence of equally valid origins at shifts of 1/2 is illuminated by the plot, as expected. (b) A similar search for the correct origin after first shifting the heavy-atom substructure by (0.15, 0.20, 0); the required origin shift is evident.