In a reliability assessment of ultrasonic time-of-flight diffraction (TOFD) inspection, probability of detection (POD) and sizing (POS) curves are developed. Experiments are performed on a complex geometry specimen with the grooved inspection surface simulating the gland seal area of a steam turbine rotor. In the reliability experiment, it is assumed and confirmed that the distribution of signal responses is normal. The effects of probe center spacing on detection and sizing are observed. The PODs developed here have a decreasing trend with flaw size which is in contrary to the generally observed increasing trend in conventional ultrasonic amplitude-based flaw sizing techniques. The reason for this decreasing POD with crack height is explained in the present study. The curves developed in this work are specific to the geometry and dimensions of the specimen with the set of notches and the probes used in the experiment. Hence, these curves can only be used under similar conditions. In TOFD inspection of similar type of complex shaped structures, e.g., turbine, the POD and POS curves developed here can be used in taking an appropriate engineering decision with respect to run, repair, or replace.