Human evaluation of solar wind data for interplanetary (IP) shock identification relies on both heuristics and pattern recognition, with the former lending itself to algorithmic representation and automation. Such detection algorithms can potentially alert forecasters of approaching shocks, providing increased warning of subsequent geomagnetic storms. However, capturing shocks with an algorithmic treatment alone is challenging, as past and present work demonstrates. We present a statistical analysis of 209 IP shocks observed at L1, and we use this information to optimize a set of shock identification criteria for use with an automated solar wind shock detection algorithm. In order to specify ranges for the threshold values used in our algorithm, we quantify discontinuities in the solar wind density, velocity, temperature, and magnetic field magnitude by analyzing 8 years of IP shocks detected by the SWEPAM and MAG instruments aboard the ACE spacecraft. Although automatic shock detection algorithms have previously been developed, in this paper we conduct a methodical optimization to refine shock identification criteria and present the optimal performance of this and similar approaches. We compute forecast skill scores for over 10,000 permutations of our shock detection criteria in order to identify the set of threshold values that yield optimal forecast skill scores. We then compare our results to previous automatic shock detection algorithms using a standard data set, and our optimized algorithm shows improvements in the reliability of automated shock detection.