The document discusses score standardization techniques for evaluating information retrieval systems. It presents two main techniques: std-CDF and std-AB. Std-CDF emphasizes moderately high and low performing systems, while std-AB provides a simple linear transformation. Std-AB is shown to have lower within-system variances and more consistent rankings across collections compared to std-CDF. The document also discusses how score standardization can enable topic set size design and comparisons across different test collections. It provides examples applying std-AB to NTCIR-12 tasks, showing it reduces variances and does not change system rankings.