The paper presents results from a computer-based assessment in which 171 early career mathematics teachers from Germany were asked to anticipate typical student errors on a given mathematical topic and identify them under time constraints. Fast and accurate perception and knowledge-based judgments are widely accepted characteristics of teacher competence. The item-wise length of anticipation time, the complexity of mathematical topics and the frequency of right or wrong given answers were used as indicators for teacher competence. The data revealed that anticipation time and the complexity of mathematical topics were related with each other. The groups of test persons with correct and incorrect answers behaved contrarily to the length of the anticipation time. Whereas test persons with correct answers needed more time to anticipate typical errors with an increasing complexity of the errors, the test persons giving false answers responded very quickly, even with an increasing error complexity. This finding confirms results of the expertise research, which emphasize that expert teachers focus more intensively if this is required by a complex task.