This study empirically determined the optimizing weight to be applied to the Wrongs Total Score in scoring rubrics of the general form = R - kW, where S is the Score, R the Rights Total, k the weight and W the Wrongs Total, if reliability is to be maximized. As is well known, the traditional formula score rests on a theoretical framework which is of dubious validity. Two instruments, variant approaches to the assessment of mathematical knowledge, were administered to approximately 1,700 entering college freshmen during an orientation period. The method consists of an iterative computer procedure for calculating split-half reliability of the tests as the weights are systematically varied throughout the region of maximization as determined by essentially canonical approaches. The results indicate that in contrast to the negative weight for the a priori formula score, a sizable positive weight maximizes reliability. The implications for rate of work as the single most reliable aspect of test performance seem clear. The validity of much educational testing rests on assumptions of fairness to those tested, achieved through optimization of standardized conditions. The study suggests that factors which alter rate-of-work characteristics of performance may be most detrimental to candidate success. (Author/DEP)