One thing that really stands out for Weaver’s career is that many so-called advanced stats hated him for much of his career and indicated that he should be just a mediocrity. But he defied the advanced stats for years and remained very successful until the last few years.
This, I think indicates a very serious flaw in the way we think about data and analytics. We expect that there is a single “good” answer and that “better” analytics would always produce something closer to that magical answer. But the better analytics, even if they might, on average, be better, will also be making different mistakes from the old. Sometimes, high FIP/low ERA combination really does mean that ERA is the better indicator of pitcher effectiveness than FIP. This raises a question, rather than an answer: when do we encounter the exceptions where we should chuck FIP in favor of ERA, not insist that FIP is still better because it is better on average and that people like Weaver must be actually bad because ERA is an empty stat. (This, of course, involves the ecological fallacy.)