An RFID reader inventories a population of tags. The reader evaluates responses from tags by categorizing them in slots. As tags are inventoried, the number of slots based on a Q-parameter is reduced. The reader addresses the tags by communicating a Q1 value for the Q parameter, generates first contents from replies received from the tags, and computes a first merit statistic based on the first contents. Then, the reader repeats the process with a Q2 value. Upon computing the first and the second merit statistics, the reader determines a Q3 value for the Q parameter. If the Q3 value is substantially equal to the Q1 value, the reader continues to receive the second replies without communicating the Q3 value. If the Q3 value is different from the Q2 value, the reader uses the Q3 value for another round of iteration heuristically converging on an optimum Q value.