Good to see better statistics being applied. However, the base methodology appears to be really weak. The presence of fixes doesn't necessarily reflect the pr

Good to see better statistics being applied. However, the base methodology appears to be really weak. The presence of fixes doesn't necessarily reflect the presence of underlying bugs: highly conscientious projects make many "fixes" that improve the software quality against an abstract ideal without directly fixing any defect, while highly unconscientious projects don't bother fixing anything. There are also many confounders. For example, I've seen project institute restrictions on refactoring changes in response to the huge amount of brownian-code-motion popular projects can receive, only to then have some contributors start misleading describing their commits, "by splitting this function up, we avoid the risk that someone would later introduce an overflow!". Considering how important software quality is to contemporary engineering, including life safety critical systems, it's disappointing that we don't see more randomize studies. E.g. take a pool of developers randomly assign them to teams using different tools each to accomplish the same task, then compare the results against each other and a hidden test suite.